1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <linux/ethtool.h> 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 52 struct netpoll_info; 53 struct device; 54 struct phy_device; 55 struct dsa_port; 56 57 struct sfp_bus; 58 /* 802.11 specific */ 59 struct wireless_dev; 60 /* 802.15.4 specific */ 61 struct wpan_dev; 62 struct mpls_dev; 63 /* UDP Tunnel offloads */ 64 struct udp_tunnel_info; 65 struct bpf_prog; 66 struct xdp_buff; 67 68 void netdev_set_default_ethtool_ops(struct net_device *dev, 69 const struct ethtool_ops *ops); 70 71 /* Backlog congestion levels */ 72 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 73 #define NET_RX_DROP 1 /* packet dropped */ 74 75 /* 76 * Transmit return codes: transmit return codes originate from three different 77 * namespaces: 78 * 79 * - qdisc return codes 80 * - driver transmit return codes 81 * - errno values 82 * 83 * Drivers are allowed to return any one of those in their hard_start_xmit() 84 * function. Real network devices commonly used with qdiscs should only return 85 * the driver transmit return codes though - when qdiscs are used, the actual 86 * transmission happens asynchronously, so the value is not propagated to 87 * higher layers. Virtual network devices transmit synchronously; in this case 88 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 89 * others are propagated to higher layers. 90 */ 91 92 /* qdisc ->enqueue() return codes. */ 93 #define NET_XMIT_SUCCESS 0x00 94 #define NET_XMIT_DROP 0x01 /* skb dropped */ 95 #define NET_XMIT_CN 0x02 /* congestion notification */ 96 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 97 98 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 99 * indicates that the device will soon be dropping packets, or already drops 100 * some packets of the same priority; prompting us to send less aggressively. */ 101 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 102 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 103 104 /* Driver transmit return codes */ 105 #define NETDEV_TX_MASK 0xf0 106 107 enum netdev_tx { 108 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 109 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 110 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 111 }; 112 typedef enum netdev_tx netdev_tx_t; 113 114 /* 115 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 116 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 117 */ 118 static inline bool dev_xmit_complete(int rc) 119 { 120 /* 121 * Positive cases with an skb consumed by a driver: 122 * - successful transmission (rc == NETDEV_TX_OK) 123 * - error while transmitting (rc < 0) 124 * - error while queueing to a different device (rc & NET_XMIT_MASK) 125 */ 126 if (likely(rc < NET_XMIT_MASK)) 127 return true; 128 129 return false; 130 } 131 132 /* 133 * Compute the worst-case header length according to the protocols 134 * used. 135 */ 136 137 #if defined(CONFIG_HYPERV_NET) 138 # define LL_MAX_HEADER 128 139 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 140 # if defined(CONFIG_MAC80211_MESH) 141 # define LL_MAX_HEADER 128 142 # else 143 # define LL_MAX_HEADER 96 144 # endif 145 #else 146 # define LL_MAX_HEADER 32 147 #endif 148 149 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 150 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 151 #define MAX_HEADER LL_MAX_HEADER 152 #else 153 #define MAX_HEADER (LL_MAX_HEADER + 48) 154 #endif 155 156 /* 157 * Old network device statistics. Fields are native words 158 * (unsigned long) so they can be read and written atomically. 159 */ 160 161 struct net_device_stats { 162 unsigned long rx_packets; 163 unsigned long tx_packets; 164 unsigned long rx_bytes; 165 unsigned long tx_bytes; 166 unsigned long rx_errors; 167 unsigned long tx_errors; 168 unsigned long rx_dropped; 169 unsigned long tx_dropped; 170 unsigned long multicast; 171 unsigned long collisions; 172 unsigned long rx_length_errors; 173 unsigned long rx_over_errors; 174 unsigned long rx_crc_errors; 175 unsigned long rx_frame_errors; 176 unsigned long rx_fifo_errors; 177 unsigned long rx_missed_errors; 178 unsigned long tx_aborted_errors; 179 unsigned long tx_carrier_errors; 180 unsigned long tx_fifo_errors; 181 unsigned long tx_heartbeat_errors; 182 unsigned long tx_window_errors; 183 unsigned long rx_compressed; 184 unsigned long tx_compressed; 185 }; 186 187 188 #include <linux/cache.h> 189 #include <linux/skbuff.h> 190 191 #ifdef CONFIG_RPS 192 #include <linux/static_key.h> 193 extern struct static_key_false rps_needed; 194 extern struct static_key_false rfs_needed; 195 #endif 196 197 struct neighbour; 198 struct neigh_parms; 199 struct sk_buff; 200 201 struct netdev_hw_addr { 202 struct list_head list; 203 unsigned char addr[MAX_ADDR_LEN]; 204 unsigned char type; 205 #define NETDEV_HW_ADDR_T_LAN 1 206 #define NETDEV_HW_ADDR_T_SAN 2 207 #define NETDEV_HW_ADDR_T_SLAVE 3 208 #define NETDEV_HW_ADDR_T_UNICAST 4 209 #define NETDEV_HW_ADDR_T_MULTICAST 5 210 bool global_use; 211 int sync_cnt; 212 int refcount; 213 int synced; 214 struct rcu_head rcu_head; 215 }; 216 217 struct netdev_hw_addr_list { 218 struct list_head list; 219 int count; 220 }; 221 222 #define netdev_hw_addr_list_count(l) ((l)->count) 223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 224 #define netdev_hw_addr_list_for_each(ha, l) \ 225 list_for_each_entry(ha, &(l)->list, list) 226 227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 229 #define netdev_for_each_uc_addr(ha, dev) \ 230 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 231 232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 234 #define netdev_for_each_mc_addr(ha, dev) \ 235 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 236 237 struct hh_cache { 238 unsigned int hh_len; 239 seqlock_t hh_lock; 240 241 /* cached hardware header; allow for machine alignment needs. */ 242 #define HH_DATA_MOD 16 243 #define HH_DATA_OFF(__len) \ 244 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 245 #define HH_DATA_ALIGN(__len) \ 246 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 247 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 248 }; 249 250 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 251 * Alternative is: 252 * dev->hard_header_len ? (dev->hard_header_len + 253 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 254 * 255 * We could use other alignment values, but we must maintain the 256 * relationship HH alignment <= LL alignment. 257 */ 258 #define LL_RESERVED_SPACE(dev) \ 259 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 260 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 261 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 262 263 struct header_ops { 264 int (*create) (struct sk_buff *skb, struct net_device *dev, 265 unsigned short type, const void *daddr, 266 const void *saddr, unsigned int len); 267 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 268 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 269 void (*cache_update)(struct hh_cache *hh, 270 const struct net_device *dev, 271 const unsigned char *haddr); 272 bool (*validate)(const char *ll_header, unsigned int len); 273 __be16 (*parse_protocol)(const struct sk_buff *skb); 274 }; 275 276 /* These flag bits are private to the generic network queueing 277 * layer; they may not be explicitly referenced by any other 278 * code. 279 */ 280 281 enum netdev_state_t { 282 __LINK_STATE_START, 283 __LINK_STATE_PRESENT, 284 __LINK_STATE_NOCARRIER, 285 __LINK_STATE_LINKWATCH_PENDING, 286 __LINK_STATE_DORMANT, 287 }; 288 289 290 /* 291 * This structure holds boot-time configured netdevice settings. They 292 * are then used in the device probing. 293 */ 294 struct netdev_boot_setup { 295 char name[IFNAMSIZ]; 296 struct ifmap map; 297 }; 298 #define NETDEV_BOOT_SETUP_MAX 8 299 300 int __init netdev_boot_setup(char *str); 301 302 struct gro_list { 303 struct list_head list; 304 int count; 305 }; 306 307 /* 308 * size of gro hash buckets, must less than bit number of 309 * napi_struct::gro_bitmask 310 */ 311 #define GRO_HASH_BUCKETS 8 312 313 /* 314 * Structure for NAPI scheduling similar to tasklet but with weighting 315 */ 316 struct napi_struct { 317 /* The poll_list must only be managed by the entity which 318 * changes the state of the NAPI_STATE_SCHED bit. This means 319 * whoever atomically sets that bit can add this napi_struct 320 * to the per-CPU poll_list, and whoever clears that bit 321 * can remove from the list right before clearing the bit. 322 */ 323 struct list_head poll_list; 324 325 unsigned long state; 326 int weight; 327 unsigned long gro_bitmask; 328 int (*poll)(struct napi_struct *, int); 329 #ifdef CONFIG_NETPOLL 330 int poll_owner; 331 #endif 332 struct net_device *dev; 333 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 334 struct sk_buff *skb; 335 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 336 int rx_count; /* length of rx_list */ 337 struct hrtimer timer; 338 struct list_head dev_list; 339 struct hlist_node napi_hash_node; 340 unsigned int napi_id; 341 }; 342 343 enum { 344 NAPI_STATE_SCHED, /* Poll is scheduled */ 345 NAPI_STATE_MISSED, /* reschedule a napi */ 346 NAPI_STATE_DISABLE, /* Disable pending */ 347 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 348 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 349 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 350 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 351 }; 352 353 enum { 354 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 355 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 356 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 357 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 358 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 359 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 360 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 361 }; 362 363 enum gro_result { 364 GRO_MERGED, 365 GRO_MERGED_FREE, 366 GRO_HELD, 367 GRO_NORMAL, 368 GRO_DROP, 369 GRO_CONSUMED, 370 }; 371 typedef enum gro_result gro_result_t; 372 373 /* 374 * enum rx_handler_result - Possible return values for rx_handlers. 375 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 376 * further. 377 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 378 * case skb->dev was changed by rx_handler. 379 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 380 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 381 * 382 * rx_handlers are functions called from inside __netif_receive_skb(), to do 383 * special processing of the skb, prior to delivery to protocol handlers. 384 * 385 * Currently, a net_device can only have a single rx_handler registered. Trying 386 * to register a second rx_handler will return -EBUSY. 387 * 388 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 389 * To unregister a rx_handler on a net_device, use 390 * netdev_rx_handler_unregister(). 391 * 392 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 393 * do with the skb. 394 * 395 * If the rx_handler consumed the skb in some way, it should return 396 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 397 * the skb to be delivered in some other way. 398 * 399 * If the rx_handler changed skb->dev, to divert the skb to another 400 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 401 * new device will be called if it exists. 402 * 403 * If the rx_handler decides the skb should be ignored, it should return 404 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 405 * are registered on exact device (ptype->dev == skb->dev). 406 * 407 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 408 * delivered, it should return RX_HANDLER_PASS. 409 * 410 * A device without a registered rx_handler will behave as if rx_handler 411 * returned RX_HANDLER_PASS. 412 */ 413 414 enum rx_handler_result { 415 RX_HANDLER_CONSUMED, 416 RX_HANDLER_ANOTHER, 417 RX_HANDLER_EXACT, 418 RX_HANDLER_PASS, 419 }; 420 typedef enum rx_handler_result rx_handler_result_t; 421 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 422 423 void __napi_schedule(struct napi_struct *n); 424 void __napi_schedule_irqoff(struct napi_struct *n); 425 426 static inline bool napi_disable_pending(struct napi_struct *n) 427 { 428 return test_bit(NAPI_STATE_DISABLE, &n->state); 429 } 430 431 bool napi_schedule_prep(struct napi_struct *n); 432 433 /** 434 * napi_schedule - schedule NAPI poll 435 * @n: NAPI context 436 * 437 * Schedule NAPI poll routine to be called if it is not already 438 * running. 439 */ 440 static inline void napi_schedule(struct napi_struct *n) 441 { 442 if (napi_schedule_prep(n)) 443 __napi_schedule(n); 444 } 445 446 /** 447 * napi_schedule_irqoff - schedule NAPI poll 448 * @n: NAPI context 449 * 450 * Variant of napi_schedule(), assuming hard irqs are masked. 451 */ 452 static inline void napi_schedule_irqoff(struct napi_struct *n) 453 { 454 if (napi_schedule_prep(n)) 455 __napi_schedule_irqoff(n); 456 } 457 458 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 459 static inline bool napi_reschedule(struct napi_struct *napi) 460 { 461 if (napi_schedule_prep(napi)) { 462 __napi_schedule(napi); 463 return true; 464 } 465 return false; 466 } 467 468 bool napi_complete_done(struct napi_struct *n, int work_done); 469 /** 470 * napi_complete - NAPI processing complete 471 * @n: NAPI context 472 * 473 * Mark NAPI processing as complete. 474 * Consider using napi_complete_done() instead. 475 * Return false if device should avoid rearming interrupts. 476 */ 477 static inline bool napi_complete(struct napi_struct *n) 478 { 479 return napi_complete_done(n, 0); 480 } 481 482 /** 483 * napi_hash_del - remove a NAPI from global table 484 * @napi: NAPI context 485 * 486 * Warning: caller must observe RCU grace period 487 * before freeing memory containing @napi, if 488 * this function returns true. 489 * Note: core networking stack automatically calls it 490 * from netif_napi_del(). 491 * Drivers might want to call this helper to combine all 492 * the needed RCU grace periods into a single one. 493 */ 494 bool napi_hash_del(struct napi_struct *napi); 495 496 /** 497 * napi_disable - prevent NAPI from scheduling 498 * @n: NAPI context 499 * 500 * Stop NAPI from being scheduled on this context. 501 * Waits till any outstanding processing completes. 502 */ 503 void napi_disable(struct napi_struct *n); 504 505 /** 506 * napi_enable - enable NAPI scheduling 507 * @n: NAPI context 508 * 509 * Resume NAPI from being scheduled on this context. 510 * Must be paired with napi_disable. 511 */ 512 static inline void napi_enable(struct napi_struct *n) 513 { 514 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 515 smp_mb__before_atomic(); 516 clear_bit(NAPI_STATE_SCHED, &n->state); 517 clear_bit(NAPI_STATE_NPSVC, &n->state); 518 } 519 520 /** 521 * napi_synchronize - wait until NAPI is not running 522 * @n: NAPI context 523 * 524 * Wait until NAPI is done being scheduled on this context. 525 * Waits till any outstanding processing completes but 526 * does not disable future activations. 527 */ 528 static inline void napi_synchronize(const struct napi_struct *n) 529 { 530 if (IS_ENABLED(CONFIG_SMP)) 531 while (test_bit(NAPI_STATE_SCHED, &n->state)) 532 msleep(1); 533 else 534 barrier(); 535 } 536 537 /** 538 * napi_if_scheduled_mark_missed - if napi is running, set the 539 * NAPIF_STATE_MISSED 540 * @n: NAPI context 541 * 542 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 543 * NAPI is scheduled. 544 **/ 545 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 546 { 547 unsigned long val, new; 548 549 do { 550 val = READ_ONCE(n->state); 551 if (val & NAPIF_STATE_DISABLE) 552 return true; 553 554 if (!(val & NAPIF_STATE_SCHED)) 555 return false; 556 557 new = val | NAPIF_STATE_MISSED; 558 } while (cmpxchg(&n->state, val, new) != val); 559 560 return true; 561 } 562 563 enum netdev_queue_state_t { 564 __QUEUE_STATE_DRV_XOFF, 565 __QUEUE_STATE_STACK_XOFF, 566 __QUEUE_STATE_FROZEN, 567 }; 568 569 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 570 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 571 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 572 573 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 574 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 575 QUEUE_STATE_FROZEN) 576 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 577 QUEUE_STATE_FROZEN) 578 579 /* 580 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 581 * netif_tx_* functions below are used to manipulate this flag. The 582 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 583 * queue independently. The netif_xmit_*stopped functions below are called 584 * to check if the queue has been stopped by the driver or stack (either 585 * of the XOFF bits are set in the state). Drivers should not need to call 586 * netif_xmit*stopped functions, they should only be using netif_tx_*. 587 */ 588 589 struct netdev_queue { 590 /* 591 * read-mostly part 592 */ 593 struct net_device *dev; 594 struct Qdisc __rcu *qdisc; 595 struct Qdisc *qdisc_sleeping; 596 #ifdef CONFIG_SYSFS 597 struct kobject kobj; 598 #endif 599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 600 int numa_node; 601 #endif 602 unsigned long tx_maxrate; 603 /* 604 * Number of TX timeouts for this queue 605 * (/sys/class/net/DEV/Q/trans_timeout) 606 */ 607 unsigned long trans_timeout; 608 609 /* Subordinate device that the queue has been assigned to */ 610 struct net_device *sb_dev; 611 #ifdef CONFIG_XDP_SOCKETS 612 struct xdp_umem *umem; 613 #endif 614 /* 615 * write-mostly part 616 */ 617 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 618 int xmit_lock_owner; 619 /* 620 * Time (in jiffies) of last Tx 621 */ 622 unsigned long trans_start; 623 624 unsigned long state; 625 626 #ifdef CONFIG_BQL 627 struct dql dql; 628 #endif 629 } ____cacheline_aligned_in_smp; 630 631 extern int sysctl_fb_tunnels_only_for_init_net; 632 extern int sysctl_devconf_inherit_init_net; 633 634 static inline bool net_has_fallback_tunnels(const struct net *net) 635 { 636 return net == &init_net || 637 !IS_ENABLED(CONFIG_SYSCTL) || 638 !sysctl_fb_tunnels_only_for_init_net; 639 } 640 641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 642 { 643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 644 return q->numa_node; 645 #else 646 return NUMA_NO_NODE; 647 #endif 648 } 649 650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 651 { 652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 653 q->numa_node = node; 654 #endif 655 } 656 657 #ifdef CONFIG_RPS 658 /* 659 * This structure holds an RPS map which can be of variable length. The 660 * map is an array of CPUs. 661 */ 662 struct rps_map { 663 unsigned int len; 664 struct rcu_head rcu; 665 u16 cpus[0]; 666 }; 667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 668 669 /* 670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 671 * tail pointer for that CPU's input queue at the time of last enqueue, and 672 * a hardware filter index. 673 */ 674 struct rps_dev_flow { 675 u16 cpu; 676 u16 filter; 677 unsigned int last_qtail; 678 }; 679 #define RPS_NO_FILTER 0xffff 680 681 /* 682 * The rps_dev_flow_table structure contains a table of flow mappings. 683 */ 684 struct rps_dev_flow_table { 685 unsigned int mask; 686 struct rcu_head rcu; 687 struct rps_dev_flow flows[0]; 688 }; 689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 690 ((_num) * sizeof(struct rps_dev_flow))) 691 692 /* 693 * The rps_sock_flow_table contains mappings of flows to the last CPU 694 * on which they were processed by the application (set in recvmsg). 695 * Each entry is a 32bit value. Upper part is the high-order bits 696 * of flow hash, lower part is CPU number. 697 * rps_cpu_mask is used to partition the space, depending on number of 698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 700 * meaning we use 32-6=26 bits for the hash. 701 */ 702 struct rps_sock_flow_table { 703 u32 mask; 704 705 u32 ents[0] ____cacheline_aligned_in_smp; 706 }; 707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 708 709 #define RPS_NO_CPU 0xffff 710 711 extern u32 rps_cpu_mask; 712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 713 714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 715 u32 hash) 716 { 717 if (table && hash) { 718 unsigned int index = hash & table->mask; 719 u32 val = hash & ~rps_cpu_mask; 720 721 /* We only give a hint, preemption can change CPU under us */ 722 val |= raw_smp_processor_id(); 723 724 if (table->ents[index] != val) 725 table->ents[index] = val; 726 } 727 } 728 729 #ifdef CONFIG_RFS_ACCEL 730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 731 u16 filter_id); 732 #endif 733 #endif /* CONFIG_RPS */ 734 735 /* This structure contains an instance of an RX queue. */ 736 struct netdev_rx_queue { 737 #ifdef CONFIG_RPS 738 struct rps_map __rcu *rps_map; 739 struct rps_dev_flow_table __rcu *rps_flow_table; 740 #endif 741 struct kobject kobj; 742 struct net_device *dev; 743 struct xdp_rxq_info xdp_rxq; 744 #ifdef CONFIG_XDP_SOCKETS 745 struct xdp_umem *umem; 746 #endif 747 } ____cacheline_aligned_in_smp; 748 749 /* 750 * RX queue sysfs structures and functions. 751 */ 752 struct rx_queue_attribute { 753 struct attribute attr; 754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 755 ssize_t (*store)(struct netdev_rx_queue *queue, 756 const char *buf, size_t len); 757 }; 758 759 #ifdef CONFIG_XPS 760 /* 761 * This structure holds an XPS map which can be of variable length. The 762 * map is an array of queues. 763 */ 764 struct xps_map { 765 unsigned int len; 766 unsigned int alloc_len; 767 struct rcu_head rcu; 768 u16 queues[0]; 769 }; 770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 772 - sizeof(struct xps_map)) / sizeof(u16)) 773 774 /* 775 * This structure holds all XPS maps for device. Maps are indexed by CPU. 776 */ 777 struct xps_dev_maps { 778 struct rcu_head rcu; 779 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */ 780 }; 781 782 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 784 785 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 786 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 787 788 #endif /* CONFIG_XPS */ 789 790 #define TC_MAX_QUEUE 16 791 #define TC_BITMASK 15 792 /* HW offloaded queuing disciplines txq count and offset maps */ 793 struct netdev_tc_txq { 794 u16 count; 795 u16 offset; 796 }; 797 798 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 799 /* 800 * This structure is to hold information about the device 801 * configured to run FCoE protocol stack. 802 */ 803 struct netdev_fcoe_hbainfo { 804 char manufacturer[64]; 805 char serial_number[64]; 806 char hardware_version[64]; 807 char driver_version[64]; 808 char optionrom_version[64]; 809 char firmware_version[64]; 810 char model[256]; 811 char model_description[256]; 812 }; 813 #endif 814 815 #define MAX_PHYS_ITEM_ID_LEN 32 816 817 /* This structure holds a unique identifier to identify some 818 * physical item (port for example) used by a netdevice. 819 */ 820 struct netdev_phys_item_id { 821 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 822 unsigned char id_len; 823 }; 824 825 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 826 struct netdev_phys_item_id *b) 827 { 828 return a->id_len == b->id_len && 829 memcmp(a->id, b->id, a->id_len) == 0; 830 } 831 832 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 833 struct sk_buff *skb, 834 struct net_device *sb_dev); 835 836 enum tc_setup_type { 837 TC_SETUP_QDISC_MQPRIO, 838 TC_SETUP_CLSU32, 839 TC_SETUP_CLSFLOWER, 840 TC_SETUP_CLSMATCHALL, 841 TC_SETUP_CLSBPF, 842 TC_SETUP_BLOCK, 843 TC_SETUP_QDISC_CBS, 844 TC_SETUP_QDISC_RED, 845 TC_SETUP_QDISC_PRIO, 846 TC_SETUP_QDISC_MQ, 847 TC_SETUP_QDISC_ETF, 848 TC_SETUP_ROOT_QDISC, 849 TC_SETUP_QDISC_GRED, 850 TC_SETUP_QDISC_TAPRIO, 851 }; 852 853 /* These structures hold the attributes of bpf state that are being passed 854 * to the netdevice through the bpf op. 855 */ 856 enum bpf_netdev_command { 857 /* Set or clear a bpf program used in the earliest stages of packet 858 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 859 * is responsible for calling bpf_prog_put on any old progs that are 860 * stored. In case of error, the callee need not release the new prog 861 * reference, but on success it takes ownership and must bpf_prog_put 862 * when it is no longer used. 863 */ 864 XDP_SETUP_PROG, 865 XDP_SETUP_PROG_HW, 866 XDP_QUERY_PROG, 867 XDP_QUERY_PROG_HW, 868 /* BPF program for offload callbacks, invoked at program load time. */ 869 BPF_OFFLOAD_MAP_ALLOC, 870 BPF_OFFLOAD_MAP_FREE, 871 XDP_SETUP_XSK_UMEM, 872 }; 873 874 struct bpf_prog_offload_ops; 875 struct netlink_ext_ack; 876 struct xdp_umem; 877 878 struct netdev_bpf { 879 enum bpf_netdev_command command; 880 union { 881 /* XDP_SETUP_PROG */ 882 struct { 883 u32 flags; 884 struct bpf_prog *prog; 885 struct netlink_ext_ack *extack; 886 }; 887 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 888 struct { 889 u32 prog_id; 890 /* flags with which program was installed */ 891 u32 prog_flags; 892 }; 893 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 894 struct { 895 struct bpf_offloaded_map *offmap; 896 }; 897 /* XDP_SETUP_XSK_UMEM */ 898 struct { 899 struct xdp_umem *umem; 900 u16 queue_id; 901 } xsk; 902 }; 903 }; 904 905 /* Flags for ndo_xsk_wakeup. */ 906 #define XDP_WAKEUP_RX (1 << 0) 907 #define XDP_WAKEUP_TX (1 << 1) 908 909 #ifdef CONFIG_XFRM_OFFLOAD 910 struct xfrmdev_ops { 911 int (*xdo_dev_state_add) (struct xfrm_state *x); 912 void (*xdo_dev_state_delete) (struct xfrm_state *x); 913 void (*xdo_dev_state_free) (struct xfrm_state *x); 914 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 915 struct xfrm_state *x); 916 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 917 }; 918 #endif 919 920 struct dev_ifalias { 921 struct rcu_head rcuhead; 922 char ifalias[]; 923 }; 924 925 struct devlink; 926 struct tlsdev_ops; 927 928 /* 929 * This structure defines the management hooks for network devices. 930 * The following hooks can be defined; unless noted otherwise, they are 931 * optional and can be filled with a null pointer. 932 * 933 * int (*ndo_init)(struct net_device *dev); 934 * This function is called once when a network device is registered. 935 * The network device can use this for any late stage initialization 936 * or semantic validation. It can fail with an error code which will 937 * be propagated back to register_netdev. 938 * 939 * void (*ndo_uninit)(struct net_device *dev); 940 * This function is called when device is unregistered or when registration 941 * fails. It is not called if init fails. 942 * 943 * int (*ndo_open)(struct net_device *dev); 944 * This function is called when a network device transitions to the up 945 * state. 946 * 947 * int (*ndo_stop)(struct net_device *dev); 948 * This function is called when a network device transitions to the down 949 * state. 950 * 951 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 952 * struct net_device *dev); 953 * Called when a packet needs to be transmitted. 954 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 955 * the queue before that can happen; it's for obsolete devices and weird 956 * corner cases, but the stack really does a non-trivial amount 957 * of useless work if you return NETDEV_TX_BUSY. 958 * Required; cannot be NULL. 959 * 960 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 961 * struct net_device *dev 962 * netdev_features_t features); 963 * Called by core transmit path to determine if device is capable of 964 * performing offload operations on a given packet. This is to give 965 * the device an opportunity to implement any restrictions that cannot 966 * be otherwise expressed by feature flags. The check is called with 967 * the set of features that the stack has calculated and it returns 968 * those the driver believes to be appropriate. 969 * 970 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 971 * struct net_device *sb_dev); 972 * Called to decide which queue to use when device supports multiple 973 * transmit queues. 974 * 975 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 976 * This function is called to allow device receiver to make 977 * changes to configuration when multicast or promiscuous is enabled. 978 * 979 * void (*ndo_set_rx_mode)(struct net_device *dev); 980 * This function is called device changes address list filtering. 981 * If driver handles unicast address filtering, it should set 982 * IFF_UNICAST_FLT in its priv_flags. 983 * 984 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 985 * This function is called when the Media Access Control address 986 * needs to be changed. If this interface is not defined, the 987 * MAC address can not be changed. 988 * 989 * int (*ndo_validate_addr)(struct net_device *dev); 990 * Test if Media Access Control address is valid for the device. 991 * 992 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 993 * Called when a user requests an ioctl which can't be handled by 994 * the generic interface code. If not defined ioctls return 995 * not supported error code. 996 * 997 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 998 * Used to set network devices bus interface parameters. This interface 999 * is retained for legacy reasons; new devices should use the bus 1000 * interface (PCI) for low level management. 1001 * 1002 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1003 * Called when a user wants to change the Maximum Transfer Unit 1004 * of a device. 1005 * 1006 * void (*ndo_tx_timeout)(struct net_device *dev); 1007 * Callback used when the transmitter has not made any progress 1008 * for dev->watchdog ticks. 1009 * 1010 * void (*ndo_get_stats64)(struct net_device *dev, 1011 * struct rtnl_link_stats64 *storage); 1012 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1013 * Called when a user wants to get the network device usage 1014 * statistics. Drivers must do one of the following: 1015 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1016 * rtnl_link_stats64 structure passed by the caller. 1017 * 2. Define @ndo_get_stats to update a net_device_stats structure 1018 * (which should normally be dev->stats) and return a pointer to 1019 * it. The structure may be changed asynchronously only if each 1020 * field is written atomically. 1021 * 3. Update dev->stats asynchronously and atomically, and define 1022 * neither operation. 1023 * 1024 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1025 * Return true if this device supports offload stats of this attr_id. 1026 * 1027 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1028 * void *attr_data) 1029 * Get statistics for offload operations by attr_id. Write it into the 1030 * attr_data pointer. 1031 * 1032 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1033 * If device supports VLAN filtering this function is called when a 1034 * VLAN id is registered. 1035 * 1036 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1037 * If device supports VLAN filtering this function is called when a 1038 * VLAN id is unregistered. 1039 * 1040 * void (*ndo_poll_controller)(struct net_device *dev); 1041 * 1042 * SR-IOV management functions. 1043 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1044 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1045 * u8 qos, __be16 proto); 1046 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1047 * int max_tx_rate); 1048 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1049 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1050 * int (*ndo_get_vf_config)(struct net_device *dev, 1051 * int vf, struct ifla_vf_info *ivf); 1052 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1053 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1054 * struct nlattr *port[]); 1055 * 1056 * Enable or disable the VF ability to query its RSS Redirection Table and 1057 * Hash Key. This is needed since on some devices VF share this information 1058 * with PF and querying it may introduce a theoretical security risk. 1059 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1060 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1061 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1062 * void *type_data); 1063 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1064 * This is always called from the stack with the rtnl lock held and netif 1065 * tx queues stopped. This allows the netdevice to perform queue 1066 * management safely. 1067 * 1068 * Fiber Channel over Ethernet (FCoE) offload functions. 1069 * int (*ndo_fcoe_enable)(struct net_device *dev); 1070 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1071 * so the underlying device can perform whatever needed configuration or 1072 * initialization to support acceleration of FCoE traffic. 1073 * 1074 * int (*ndo_fcoe_disable)(struct net_device *dev); 1075 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1076 * so the underlying device can perform whatever needed clean-ups to 1077 * stop supporting acceleration of FCoE traffic. 1078 * 1079 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1080 * struct scatterlist *sgl, unsigned int sgc); 1081 * Called when the FCoE Initiator wants to initialize an I/O that 1082 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1083 * perform necessary setup and returns 1 to indicate the device is set up 1084 * successfully to perform DDP on this I/O, otherwise this returns 0. 1085 * 1086 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1087 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1088 * indicated by the FC exchange id 'xid', so the underlying device can 1089 * clean up and reuse resources for later DDP requests. 1090 * 1091 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1092 * struct scatterlist *sgl, unsigned int sgc); 1093 * Called when the FCoE Target wants to initialize an I/O that 1094 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1095 * perform necessary setup and returns 1 to indicate the device is set up 1096 * successfully to perform DDP on this I/O, otherwise this returns 0. 1097 * 1098 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1099 * struct netdev_fcoe_hbainfo *hbainfo); 1100 * Called when the FCoE Protocol stack wants information on the underlying 1101 * device. This information is utilized by the FCoE protocol stack to 1102 * register attributes with Fiber Channel management service as per the 1103 * FC-GS Fabric Device Management Information(FDMI) specification. 1104 * 1105 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1106 * Called when the underlying device wants to override default World Wide 1107 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1108 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1109 * protocol stack to use. 1110 * 1111 * RFS acceleration. 1112 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1113 * u16 rxq_index, u32 flow_id); 1114 * Set hardware filter for RFS. rxq_index is the target queue index; 1115 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1116 * Return the filter ID on success, or a negative error code. 1117 * 1118 * Slave management functions (for bridge, bonding, etc). 1119 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1120 * Called to make another netdev an underling. 1121 * 1122 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1123 * Called to release previously enslaved netdev. 1124 * 1125 * Feature/offload setting functions. 1126 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1127 * netdev_features_t features); 1128 * Adjusts the requested feature flags according to device-specific 1129 * constraints, and returns the resulting flags. Must not modify 1130 * the device state. 1131 * 1132 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1133 * Called to update device configuration to new features. Passed 1134 * feature set might be less than what was returned by ndo_fix_features()). 1135 * Must return >0 or -errno if it changed dev->features itself. 1136 * 1137 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1138 * struct net_device *dev, 1139 * const unsigned char *addr, u16 vid, u16 flags, 1140 * struct netlink_ext_ack *extack); 1141 * Adds an FDB entry to dev for addr. 1142 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1143 * struct net_device *dev, 1144 * const unsigned char *addr, u16 vid) 1145 * Deletes the FDB entry from dev coresponding to addr. 1146 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1147 * struct net_device *dev, struct net_device *filter_dev, 1148 * int *idx) 1149 * Used to add FDB entries to dump requests. Implementers should add 1150 * entries to skb and update idx with the number of entries. 1151 * 1152 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1153 * u16 flags, struct netlink_ext_ack *extack) 1154 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1155 * struct net_device *dev, u32 filter_mask, 1156 * int nlflags) 1157 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1158 * u16 flags); 1159 * 1160 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1161 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1162 * which do not represent real hardware may define this to allow their 1163 * userspace components to manage their virtual carrier state. Devices 1164 * that determine carrier state from physical hardware properties (eg 1165 * network cables) or protocol-dependent mechanisms (eg 1166 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1167 * 1168 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1169 * struct netdev_phys_item_id *ppid); 1170 * Called to get ID of physical port of this device. If driver does 1171 * not implement this, it is assumed that the hw is not able to have 1172 * multiple net devices on single physical port. 1173 * 1174 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1175 * struct netdev_phys_item_id *ppid) 1176 * Called to get the parent ID of the physical port of this device. 1177 * 1178 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1179 * struct udp_tunnel_info *ti); 1180 * Called by UDP tunnel to notify a driver about the UDP port and socket 1181 * address family that a UDP tunnel is listnening to. It is called only 1182 * when a new port starts listening. The operation is protected by the 1183 * RTNL. 1184 * 1185 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1186 * struct udp_tunnel_info *ti); 1187 * Called by UDP tunnel to notify the driver about a UDP port and socket 1188 * address family that the UDP tunnel is not listening to anymore. The 1189 * operation is protected by the RTNL. 1190 * 1191 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1192 * struct net_device *dev) 1193 * Called by upper layer devices to accelerate switching or other 1194 * station functionality into hardware. 'pdev is the lowerdev 1195 * to use for the offload and 'dev' is the net device that will 1196 * back the offload. Returns a pointer to the private structure 1197 * the upper layer will maintain. 1198 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1199 * Called by upper layer device to delete the station created 1200 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1201 * the station and priv is the structure returned by the add 1202 * operation. 1203 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1204 * int queue_index, u32 maxrate); 1205 * Called when a user wants to set a max-rate limitation of specific 1206 * TX queue. 1207 * int (*ndo_get_iflink)(const struct net_device *dev); 1208 * Called to get the iflink value of this device. 1209 * void (*ndo_change_proto_down)(struct net_device *dev, 1210 * bool proto_down); 1211 * This function is used to pass protocol port error state information 1212 * to the switch driver. The switch driver can react to the proto_down 1213 * by doing a phys down on the associated switch port. 1214 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1215 * This function is used to get egress tunnel information for given skb. 1216 * This is useful for retrieving outer tunnel header parameters while 1217 * sampling packet. 1218 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1219 * This function is used to specify the headroom that the skb must 1220 * consider when allocation skb during packet reception. Setting 1221 * appropriate rx headroom value allows avoiding skb head copy on 1222 * forward. Setting a negative value resets the rx headroom to the 1223 * default value. 1224 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1225 * This function is used to set or query state related to XDP on the 1226 * netdevice and manage BPF offload. See definition of 1227 * enum bpf_netdev_command for details. 1228 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1229 * u32 flags); 1230 * This function is used to submit @n XDP packets for transmit on a 1231 * netdevice. Returns number of frames successfully transmitted, frames 1232 * that got dropped are freed/returned via xdp_return_frame(). 1233 * Returns negative number, means general error invoking ndo, meaning 1234 * no frames were xmit'ed and core-caller will free all frames. 1235 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1236 * This function is used to wake up the softirq, ksoftirqd or kthread 1237 * responsible for sending and/or receiving packets on a specific 1238 * queue id bound to an AF_XDP socket. The flags field specifies if 1239 * only RX, only Tx, or both should be woken up using the flags 1240 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1241 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1242 * Get devlink port instance associated with a given netdev. 1243 * Called with a reference on the netdevice and devlink locks only, 1244 * rtnl_lock is not held. 1245 */ 1246 struct net_device_ops { 1247 int (*ndo_init)(struct net_device *dev); 1248 void (*ndo_uninit)(struct net_device *dev); 1249 int (*ndo_open)(struct net_device *dev); 1250 int (*ndo_stop)(struct net_device *dev); 1251 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1252 struct net_device *dev); 1253 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1254 struct net_device *dev, 1255 netdev_features_t features); 1256 u16 (*ndo_select_queue)(struct net_device *dev, 1257 struct sk_buff *skb, 1258 struct net_device *sb_dev); 1259 void (*ndo_change_rx_flags)(struct net_device *dev, 1260 int flags); 1261 void (*ndo_set_rx_mode)(struct net_device *dev); 1262 int (*ndo_set_mac_address)(struct net_device *dev, 1263 void *addr); 1264 int (*ndo_validate_addr)(struct net_device *dev); 1265 int (*ndo_do_ioctl)(struct net_device *dev, 1266 struct ifreq *ifr, int cmd); 1267 int (*ndo_set_config)(struct net_device *dev, 1268 struct ifmap *map); 1269 int (*ndo_change_mtu)(struct net_device *dev, 1270 int new_mtu); 1271 int (*ndo_neigh_setup)(struct net_device *dev, 1272 struct neigh_parms *); 1273 void (*ndo_tx_timeout) (struct net_device *dev); 1274 1275 void (*ndo_get_stats64)(struct net_device *dev, 1276 struct rtnl_link_stats64 *storage); 1277 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1278 int (*ndo_get_offload_stats)(int attr_id, 1279 const struct net_device *dev, 1280 void *attr_data); 1281 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1282 1283 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1284 __be16 proto, u16 vid); 1285 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1286 __be16 proto, u16 vid); 1287 #ifdef CONFIG_NET_POLL_CONTROLLER 1288 void (*ndo_poll_controller)(struct net_device *dev); 1289 int (*ndo_netpoll_setup)(struct net_device *dev, 1290 struct netpoll_info *info); 1291 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1292 #endif 1293 int (*ndo_set_vf_mac)(struct net_device *dev, 1294 int queue, u8 *mac); 1295 int (*ndo_set_vf_vlan)(struct net_device *dev, 1296 int queue, u16 vlan, 1297 u8 qos, __be16 proto); 1298 int (*ndo_set_vf_rate)(struct net_device *dev, 1299 int vf, int min_tx_rate, 1300 int max_tx_rate); 1301 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1302 int vf, bool setting); 1303 int (*ndo_set_vf_trust)(struct net_device *dev, 1304 int vf, bool setting); 1305 int (*ndo_get_vf_config)(struct net_device *dev, 1306 int vf, 1307 struct ifla_vf_info *ivf); 1308 int (*ndo_set_vf_link_state)(struct net_device *dev, 1309 int vf, int link_state); 1310 int (*ndo_get_vf_stats)(struct net_device *dev, 1311 int vf, 1312 struct ifla_vf_stats 1313 *vf_stats); 1314 int (*ndo_set_vf_port)(struct net_device *dev, 1315 int vf, 1316 struct nlattr *port[]); 1317 int (*ndo_get_vf_port)(struct net_device *dev, 1318 int vf, struct sk_buff *skb); 1319 int (*ndo_set_vf_guid)(struct net_device *dev, 1320 int vf, u64 guid, 1321 int guid_type); 1322 int (*ndo_set_vf_rss_query_en)( 1323 struct net_device *dev, 1324 int vf, bool setting); 1325 int (*ndo_setup_tc)(struct net_device *dev, 1326 enum tc_setup_type type, 1327 void *type_data); 1328 #if IS_ENABLED(CONFIG_FCOE) 1329 int (*ndo_fcoe_enable)(struct net_device *dev); 1330 int (*ndo_fcoe_disable)(struct net_device *dev); 1331 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1332 u16 xid, 1333 struct scatterlist *sgl, 1334 unsigned int sgc); 1335 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1336 u16 xid); 1337 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1338 u16 xid, 1339 struct scatterlist *sgl, 1340 unsigned int sgc); 1341 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1342 struct netdev_fcoe_hbainfo *hbainfo); 1343 #endif 1344 1345 #if IS_ENABLED(CONFIG_LIBFCOE) 1346 #define NETDEV_FCOE_WWNN 0 1347 #define NETDEV_FCOE_WWPN 1 1348 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1349 u64 *wwn, int type); 1350 #endif 1351 1352 #ifdef CONFIG_RFS_ACCEL 1353 int (*ndo_rx_flow_steer)(struct net_device *dev, 1354 const struct sk_buff *skb, 1355 u16 rxq_index, 1356 u32 flow_id); 1357 #endif 1358 int (*ndo_add_slave)(struct net_device *dev, 1359 struct net_device *slave_dev, 1360 struct netlink_ext_ack *extack); 1361 int (*ndo_del_slave)(struct net_device *dev, 1362 struct net_device *slave_dev); 1363 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1364 netdev_features_t features); 1365 int (*ndo_set_features)(struct net_device *dev, 1366 netdev_features_t features); 1367 int (*ndo_neigh_construct)(struct net_device *dev, 1368 struct neighbour *n); 1369 void (*ndo_neigh_destroy)(struct net_device *dev, 1370 struct neighbour *n); 1371 1372 int (*ndo_fdb_add)(struct ndmsg *ndm, 1373 struct nlattr *tb[], 1374 struct net_device *dev, 1375 const unsigned char *addr, 1376 u16 vid, 1377 u16 flags, 1378 struct netlink_ext_ack *extack); 1379 int (*ndo_fdb_del)(struct ndmsg *ndm, 1380 struct nlattr *tb[], 1381 struct net_device *dev, 1382 const unsigned char *addr, 1383 u16 vid); 1384 int (*ndo_fdb_dump)(struct sk_buff *skb, 1385 struct netlink_callback *cb, 1386 struct net_device *dev, 1387 struct net_device *filter_dev, 1388 int *idx); 1389 int (*ndo_fdb_get)(struct sk_buff *skb, 1390 struct nlattr *tb[], 1391 struct net_device *dev, 1392 const unsigned char *addr, 1393 u16 vid, u32 portid, u32 seq, 1394 struct netlink_ext_ack *extack); 1395 int (*ndo_bridge_setlink)(struct net_device *dev, 1396 struct nlmsghdr *nlh, 1397 u16 flags, 1398 struct netlink_ext_ack *extack); 1399 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1400 u32 pid, u32 seq, 1401 struct net_device *dev, 1402 u32 filter_mask, 1403 int nlflags); 1404 int (*ndo_bridge_dellink)(struct net_device *dev, 1405 struct nlmsghdr *nlh, 1406 u16 flags); 1407 int (*ndo_change_carrier)(struct net_device *dev, 1408 bool new_carrier); 1409 int (*ndo_get_phys_port_id)(struct net_device *dev, 1410 struct netdev_phys_item_id *ppid); 1411 int (*ndo_get_port_parent_id)(struct net_device *dev, 1412 struct netdev_phys_item_id *ppid); 1413 int (*ndo_get_phys_port_name)(struct net_device *dev, 1414 char *name, size_t len); 1415 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1416 struct udp_tunnel_info *ti); 1417 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1418 struct udp_tunnel_info *ti); 1419 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1420 struct net_device *dev); 1421 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1422 void *priv); 1423 1424 int (*ndo_get_lock_subclass)(struct net_device *dev); 1425 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1426 int queue_index, 1427 u32 maxrate); 1428 int (*ndo_get_iflink)(const struct net_device *dev); 1429 int (*ndo_change_proto_down)(struct net_device *dev, 1430 bool proto_down); 1431 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1432 struct sk_buff *skb); 1433 void (*ndo_set_rx_headroom)(struct net_device *dev, 1434 int needed_headroom); 1435 int (*ndo_bpf)(struct net_device *dev, 1436 struct netdev_bpf *bpf); 1437 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1438 struct xdp_frame **xdp, 1439 u32 flags); 1440 int (*ndo_xsk_wakeup)(struct net_device *dev, 1441 u32 queue_id, u32 flags); 1442 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1443 }; 1444 1445 /** 1446 * enum net_device_priv_flags - &struct net_device priv_flags 1447 * 1448 * These are the &struct net_device, they are only set internally 1449 * by drivers and used in the kernel. These flags are invisible to 1450 * userspace; this means that the order of these flags can change 1451 * during any kernel release. 1452 * 1453 * You should have a pretty good reason to be extending these flags. 1454 * 1455 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1456 * @IFF_EBRIDGE: Ethernet bridging device 1457 * @IFF_BONDING: bonding master or slave 1458 * @IFF_ISATAP: ISATAP interface (RFC4214) 1459 * @IFF_WAN_HDLC: WAN HDLC device 1460 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1461 * release skb->dst 1462 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1463 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1464 * @IFF_MACVLAN_PORT: device used as macvlan port 1465 * @IFF_BRIDGE_PORT: device used as bridge port 1466 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1467 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1468 * @IFF_UNICAST_FLT: Supports unicast filtering 1469 * @IFF_TEAM_PORT: device used as team port 1470 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1471 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1472 * change when it's running 1473 * @IFF_MACVLAN: Macvlan device 1474 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1475 * underlying stacked devices 1476 * @IFF_L3MDEV_MASTER: device is an L3 master device 1477 * @IFF_NO_QUEUE: device can run without qdisc attached 1478 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1479 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1480 * @IFF_TEAM: device is a team device 1481 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1482 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1483 * entity (i.e. the master device for bridged veth) 1484 * @IFF_MACSEC: device is a MACsec device 1485 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1486 * @IFF_FAILOVER: device is a failover master device 1487 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1488 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1489 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1490 */ 1491 enum netdev_priv_flags { 1492 IFF_802_1Q_VLAN = 1<<0, 1493 IFF_EBRIDGE = 1<<1, 1494 IFF_BONDING = 1<<2, 1495 IFF_ISATAP = 1<<3, 1496 IFF_WAN_HDLC = 1<<4, 1497 IFF_XMIT_DST_RELEASE = 1<<5, 1498 IFF_DONT_BRIDGE = 1<<6, 1499 IFF_DISABLE_NETPOLL = 1<<7, 1500 IFF_MACVLAN_PORT = 1<<8, 1501 IFF_BRIDGE_PORT = 1<<9, 1502 IFF_OVS_DATAPATH = 1<<10, 1503 IFF_TX_SKB_SHARING = 1<<11, 1504 IFF_UNICAST_FLT = 1<<12, 1505 IFF_TEAM_PORT = 1<<13, 1506 IFF_SUPP_NOFCS = 1<<14, 1507 IFF_LIVE_ADDR_CHANGE = 1<<15, 1508 IFF_MACVLAN = 1<<16, 1509 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1510 IFF_L3MDEV_MASTER = 1<<18, 1511 IFF_NO_QUEUE = 1<<19, 1512 IFF_OPENVSWITCH = 1<<20, 1513 IFF_L3MDEV_SLAVE = 1<<21, 1514 IFF_TEAM = 1<<22, 1515 IFF_RXFH_CONFIGURED = 1<<23, 1516 IFF_PHONY_HEADROOM = 1<<24, 1517 IFF_MACSEC = 1<<25, 1518 IFF_NO_RX_HANDLER = 1<<26, 1519 IFF_FAILOVER = 1<<27, 1520 IFF_FAILOVER_SLAVE = 1<<28, 1521 IFF_L3MDEV_RX_HANDLER = 1<<29, 1522 IFF_LIVE_RENAME_OK = 1<<30, 1523 }; 1524 1525 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1526 #define IFF_EBRIDGE IFF_EBRIDGE 1527 #define IFF_BONDING IFF_BONDING 1528 #define IFF_ISATAP IFF_ISATAP 1529 #define IFF_WAN_HDLC IFF_WAN_HDLC 1530 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1531 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1532 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1533 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1534 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1535 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1536 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1537 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1538 #define IFF_TEAM_PORT IFF_TEAM_PORT 1539 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1540 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1541 #define IFF_MACVLAN IFF_MACVLAN 1542 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1543 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1544 #define IFF_NO_QUEUE IFF_NO_QUEUE 1545 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1546 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1547 #define IFF_TEAM IFF_TEAM 1548 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1549 #define IFF_MACSEC IFF_MACSEC 1550 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1551 #define IFF_FAILOVER IFF_FAILOVER 1552 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1553 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1554 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1555 1556 /** 1557 * struct net_device - The DEVICE structure. 1558 * 1559 * Actually, this whole structure is a big mistake. It mixes I/O 1560 * data with strictly "high-level" data, and it has to know about 1561 * almost every data structure used in the INET module. 1562 * 1563 * @name: This is the first field of the "visible" part of this structure 1564 * (i.e. as seen by users in the "Space.c" file). It is the name 1565 * of the interface. 1566 * 1567 * @name_hlist: Device name hash chain, please keep it close to name[] 1568 * @ifalias: SNMP alias 1569 * @mem_end: Shared memory end 1570 * @mem_start: Shared memory start 1571 * @base_addr: Device I/O address 1572 * @irq: Device IRQ number 1573 * 1574 * @state: Generic network queuing layer state, see netdev_state_t 1575 * @dev_list: The global list of network devices 1576 * @napi_list: List entry used for polling NAPI devices 1577 * @unreg_list: List entry when we are unregistering the 1578 * device; see the function unregister_netdev 1579 * @close_list: List entry used when we are closing the device 1580 * @ptype_all: Device-specific packet handlers for all protocols 1581 * @ptype_specific: Device-specific, protocol-specific packet handlers 1582 * 1583 * @adj_list: Directly linked devices, like slaves for bonding 1584 * @features: Currently active device features 1585 * @hw_features: User-changeable features 1586 * 1587 * @wanted_features: User-requested features 1588 * @vlan_features: Mask of features inheritable by VLAN devices 1589 * 1590 * @hw_enc_features: Mask of features inherited by encapsulating devices 1591 * This field indicates what encapsulation 1592 * offloads the hardware is capable of doing, 1593 * and drivers will need to set them appropriately. 1594 * 1595 * @mpls_features: Mask of features inheritable by MPLS 1596 * 1597 * @ifindex: interface index 1598 * @group: The group the device belongs to 1599 * 1600 * @stats: Statistics struct, which was left as a legacy, use 1601 * rtnl_link_stats64 instead 1602 * 1603 * @rx_dropped: Dropped packets by core network, 1604 * do not use this in drivers 1605 * @tx_dropped: Dropped packets by core network, 1606 * do not use this in drivers 1607 * @rx_nohandler: nohandler dropped packets by core network on 1608 * inactive devices, do not use this in drivers 1609 * @carrier_up_count: Number of times the carrier has been up 1610 * @carrier_down_count: Number of times the carrier has been down 1611 * 1612 * @wireless_handlers: List of functions to handle Wireless Extensions, 1613 * instead of ioctl, 1614 * see <net/iw_handler.h> for details. 1615 * @wireless_data: Instance data managed by the core of wireless extensions 1616 * 1617 * @netdev_ops: Includes several pointers to callbacks, 1618 * if one wants to override the ndo_*() functions 1619 * @ethtool_ops: Management operations 1620 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1621 * discovery handling. Necessary for e.g. 6LoWPAN. 1622 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1623 * of Layer 2 headers. 1624 * 1625 * @flags: Interface flags (a la BSD) 1626 * @priv_flags: Like 'flags' but invisible to userspace, 1627 * see if.h for the definitions 1628 * @gflags: Global flags ( kept as legacy ) 1629 * @padded: How much padding added by alloc_netdev() 1630 * @operstate: RFC2863 operstate 1631 * @link_mode: Mapping policy to operstate 1632 * @if_port: Selectable AUI, TP, ... 1633 * @dma: DMA channel 1634 * @mtu: Interface MTU value 1635 * @min_mtu: Interface Minimum MTU value 1636 * @max_mtu: Interface Maximum MTU value 1637 * @type: Interface hardware type 1638 * @hard_header_len: Maximum hardware header length. 1639 * @min_header_len: Minimum hardware header length 1640 * 1641 * @needed_headroom: Extra headroom the hardware may need, but not in all 1642 * cases can this be guaranteed 1643 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1644 * cases can this be guaranteed. Some cases also use 1645 * LL_MAX_HEADER instead to allocate the skb 1646 * 1647 * interface address info: 1648 * 1649 * @perm_addr: Permanent hw address 1650 * @addr_assign_type: Hw address assignment type 1651 * @addr_len: Hardware address length 1652 * @neigh_priv_len: Used in neigh_alloc() 1653 * @dev_id: Used to differentiate devices that share 1654 * the same link layer address 1655 * @dev_port: Used to differentiate devices that share 1656 * the same function 1657 * @addr_list_lock: XXX: need comments on this one 1658 * @uc_promisc: Counter that indicates promiscuous mode 1659 * has been enabled due to the need to listen to 1660 * additional unicast addresses in a device that 1661 * does not implement ndo_set_rx_mode() 1662 * @uc: unicast mac addresses 1663 * @mc: multicast mac addresses 1664 * @dev_addrs: list of device hw addresses 1665 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1666 * @promiscuity: Number of times the NIC is told to work in 1667 * promiscuous mode; if it becomes 0 the NIC will 1668 * exit promiscuous mode 1669 * @allmulti: Counter, enables or disables allmulticast mode 1670 * 1671 * @vlan_info: VLAN info 1672 * @dsa_ptr: dsa specific data 1673 * @tipc_ptr: TIPC specific data 1674 * @atalk_ptr: AppleTalk link 1675 * @ip_ptr: IPv4 specific data 1676 * @dn_ptr: DECnet specific data 1677 * @ip6_ptr: IPv6 specific data 1678 * @ax25_ptr: AX.25 specific data 1679 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1680 * 1681 * @dev_addr: Hw address (before bcast, 1682 * because most packets are unicast) 1683 * 1684 * @_rx: Array of RX queues 1685 * @num_rx_queues: Number of RX queues 1686 * allocated at register_netdev() time 1687 * @real_num_rx_queues: Number of RX queues currently active in device 1688 * 1689 * @rx_handler: handler for received packets 1690 * @rx_handler_data: XXX: need comments on this one 1691 * @miniq_ingress: ingress/clsact qdisc specific data for 1692 * ingress processing 1693 * @ingress_queue: XXX: need comments on this one 1694 * @broadcast: hw bcast address 1695 * 1696 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1697 * indexed by RX queue number. Assigned by driver. 1698 * This must only be set if the ndo_rx_flow_steer 1699 * operation is defined 1700 * @index_hlist: Device index hash chain 1701 * 1702 * @_tx: Array of TX queues 1703 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1704 * @real_num_tx_queues: Number of TX queues currently active in device 1705 * @qdisc: Root qdisc from userspace point of view 1706 * @tx_queue_len: Max frames per queue allowed 1707 * @tx_global_lock: XXX: need comments on this one 1708 * 1709 * @xps_maps: XXX: need comments on this one 1710 * @miniq_egress: clsact qdisc specific data for 1711 * egress processing 1712 * @watchdog_timeo: Represents the timeout that is used by 1713 * the watchdog (see dev_watchdog()) 1714 * @watchdog_timer: List of timers 1715 * 1716 * @pcpu_refcnt: Number of references to this device 1717 * @todo_list: Delayed register/unregister 1718 * @link_watch_list: XXX: need comments on this one 1719 * 1720 * @reg_state: Register/unregister state machine 1721 * @dismantle: Device is going to be freed 1722 * @rtnl_link_state: This enum represents the phases of creating 1723 * a new link 1724 * 1725 * @needs_free_netdev: Should unregister perform free_netdev? 1726 * @priv_destructor: Called from unregister 1727 * @npinfo: XXX: need comments on this one 1728 * @nd_net: Network namespace this network device is inside 1729 * 1730 * @ml_priv: Mid-layer private 1731 * @lstats: Loopback statistics 1732 * @tstats: Tunnel statistics 1733 * @dstats: Dummy statistics 1734 * @vstats: Virtual ethernet statistics 1735 * 1736 * @garp_port: GARP 1737 * @mrp_port: MRP 1738 * 1739 * @dev: Class/net/name entry 1740 * @sysfs_groups: Space for optional device, statistics and wireless 1741 * sysfs groups 1742 * 1743 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1744 * @rtnl_link_ops: Rtnl_link_ops 1745 * 1746 * @gso_max_size: Maximum size of generic segmentation offload 1747 * @gso_max_segs: Maximum number of segments that can be passed to the 1748 * NIC for GSO 1749 * 1750 * @dcbnl_ops: Data Center Bridging netlink ops 1751 * @num_tc: Number of traffic classes in the net device 1752 * @tc_to_txq: XXX: need comments on this one 1753 * @prio_tc_map: XXX: need comments on this one 1754 * 1755 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1756 * 1757 * @priomap: XXX: need comments on this one 1758 * @phydev: Physical device may attach itself 1759 * for hardware timestamping 1760 * @sfp_bus: attached &struct sfp_bus structure. 1761 * 1762 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1763 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1764 * 1765 * @proto_down: protocol port state information can be sent to the 1766 * switch driver and used to set the phys state of the 1767 * switch port. 1768 * 1769 * @wol_enabled: Wake-on-LAN is enabled 1770 * 1771 * FIXME: cleanup struct net_device such that network protocol info 1772 * moves out. 1773 */ 1774 1775 struct net_device { 1776 char name[IFNAMSIZ]; 1777 struct hlist_node name_hlist; 1778 struct dev_ifalias __rcu *ifalias; 1779 /* 1780 * I/O specific fields 1781 * FIXME: Merge these and struct ifmap into one 1782 */ 1783 unsigned long mem_end; 1784 unsigned long mem_start; 1785 unsigned long base_addr; 1786 int irq; 1787 1788 /* 1789 * Some hardware also needs these fields (state,dev_list, 1790 * napi_list,unreg_list,close_list) but they are not 1791 * part of the usual set specified in Space.c. 1792 */ 1793 1794 unsigned long state; 1795 1796 struct list_head dev_list; 1797 struct list_head napi_list; 1798 struct list_head unreg_list; 1799 struct list_head close_list; 1800 struct list_head ptype_all; 1801 struct list_head ptype_specific; 1802 1803 struct { 1804 struct list_head upper; 1805 struct list_head lower; 1806 } adj_list; 1807 1808 netdev_features_t features; 1809 netdev_features_t hw_features; 1810 netdev_features_t wanted_features; 1811 netdev_features_t vlan_features; 1812 netdev_features_t hw_enc_features; 1813 netdev_features_t mpls_features; 1814 netdev_features_t gso_partial_features; 1815 1816 int ifindex; 1817 int group; 1818 1819 struct net_device_stats stats; 1820 1821 atomic_long_t rx_dropped; 1822 atomic_long_t tx_dropped; 1823 atomic_long_t rx_nohandler; 1824 1825 /* Stats to monitor link on/off, flapping */ 1826 atomic_t carrier_up_count; 1827 atomic_t carrier_down_count; 1828 1829 #ifdef CONFIG_WIRELESS_EXT 1830 const struct iw_handler_def *wireless_handlers; 1831 struct iw_public_data *wireless_data; 1832 #endif 1833 const struct net_device_ops *netdev_ops; 1834 const struct ethtool_ops *ethtool_ops; 1835 #ifdef CONFIG_NET_L3_MASTER_DEV 1836 const struct l3mdev_ops *l3mdev_ops; 1837 #endif 1838 #if IS_ENABLED(CONFIG_IPV6) 1839 const struct ndisc_ops *ndisc_ops; 1840 #endif 1841 1842 #ifdef CONFIG_XFRM_OFFLOAD 1843 const struct xfrmdev_ops *xfrmdev_ops; 1844 #endif 1845 1846 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1847 const struct tlsdev_ops *tlsdev_ops; 1848 #endif 1849 1850 const struct header_ops *header_ops; 1851 1852 unsigned int flags; 1853 unsigned int priv_flags; 1854 1855 unsigned short gflags; 1856 unsigned short padded; 1857 1858 unsigned char operstate; 1859 unsigned char link_mode; 1860 1861 unsigned char if_port; 1862 unsigned char dma; 1863 1864 unsigned int mtu; 1865 unsigned int min_mtu; 1866 unsigned int max_mtu; 1867 unsigned short type; 1868 unsigned short hard_header_len; 1869 unsigned char min_header_len; 1870 1871 unsigned short needed_headroom; 1872 unsigned short needed_tailroom; 1873 1874 /* Interface address info. */ 1875 unsigned char perm_addr[MAX_ADDR_LEN]; 1876 unsigned char addr_assign_type; 1877 unsigned char addr_len; 1878 unsigned short neigh_priv_len; 1879 unsigned short dev_id; 1880 unsigned short dev_port; 1881 spinlock_t addr_list_lock; 1882 unsigned char name_assign_type; 1883 bool uc_promisc; 1884 struct netdev_hw_addr_list uc; 1885 struct netdev_hw_addr_list mc; 1886 struct netdev_hw_addr_list dev_addrs; 1887 1888 #ifdef CONFIG_SYSFS 1889 struct kset *queues_kset; 1890 #endif 1891 unsigned int promiscuity; 1892 unsigned int allmulti; 1893 1894 1895 /* Protocol-specific pointers */ 1896 1897 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1898 struct vlan_info __rcu *vlan_info; 1899 #endif 1900 #if IS_ENABLED(CONFIG_NET_DSA) 1901 struct dsa_port *dsa_ptr; 1902 #endif 1903 #if IS_ENABLED(CONFIG_TIPC) 1904 struct tipc_bearer __rcu *tipc_ptr; 1905 #endif 1906 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1907 void *atalk_ptr; 1908 #endif 1909 struct in_device __rcu *ip_ptr; 1910 #if IS_ENABLED(CONFIG_DECNET) 1911 struct dn_dev __rcu *dn_ptr; 1912 #endif 1913 struct inet6_dev __rcu *ip6_ptr; 1914 #if IS_ENABLED(CONFIG_AX25) 1915 void *ax25_ptr; 1916 #endif 1917 struct wireless_dev *ieee80211_ptr; 1918 struct wpan_dev *ieee802154_ptr; 1919 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1920 struct mpls_dev __rcu *mpls_ptr; 1921 #endif 1922 1923 /* 1924 * Cache lines mostly used on receive path (including eth_type_trans()) 1925 */ 1926 /* Interface address info used in eth_type_trans() */ 1927 unsigned char *dev_addr; 1928 1929 struct netdev_rx_queue *_rx; 1930 unsigned int num_rx_queues; 1931 unsigned int real_num_rx_queues; 1932 1933 struct bpf_prog __rcu *xdp_prog; 1934 unsigned long gro_flush_timeout; 1935 rx_handler_func_t __rcu *rx_handler; 1936 void __rcu *rx_handler_data; 1937 1938 #ifdef CONFIG_NET_CLS_ACT 1939 struct mini_Qdisc __rcu *miniq_ingress; 1940 #endif 1941 struct netdev_queue __rcu *ingress_queue; 1942 #ifdef CONFIG_NETFILTER_INGRESS 1943 struct nf_hook_entries __rcu *nf_hooks_ingress; 1944 #endif 1945 1946 unsigned char broadcast[MAX_ADDR_LEN]; 1947 #ifdef CONFIG_RFS_ACCEL 1948 struct cpu_rmap *rx_cpu_rmap; 1949 #endif 1950 struct hlist_node index_hlist; 1951 1952 /* 1953 * Cache lines mostly used on transmit path 1954 */ 1955 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1956 unsigned int num_tx_queues; 1957 unsigned int real_num_tx_queues; 1958 struct Qdisc *qdisc; 1959 #ifdef CONFIG_NET_SCHED 1960 DECLARE_HASHTABLE (qdisc_hash, 4); 1961 #endif 1962 unsigned int tx_queue_len; 1963 spinlock_t tx_global_lock; 1964 int watchdog_timeo; 1965 1966 #ifdef CONFIG_XPS 1967 struct xps_dev_maps __rcu *xps_cpus_map; 1968 struct xps_dev_maps __rcu *xps_rxqs_map; 1969 #endif 1970 #ifdef CONFIG_NET_CLS_ACT 1971 struct mini_Qdisc __rcu *miniq_egress; 1972 #endif 1973 1974 /* These may be needed for future network-power-down code. */ 1975 struct timer_list watchdog_timer; 1976 1977 int __percpu *pcpu_refcnt; 1978 struct list_head todo_list; 1979 1980 struct list_head link_watch_list; 1981 1982 enum { NETREG_UNINITIALIZED=0, 1983 NETREG_REGISTERED, /* completed register_netdevice */ 1984 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1985 NETREG_UNREGISTERED, /* completed unregister todo */ 1986 NETREG_RELEASED, /* called free_netdev */ 1987 NETREG_DUMMY, /* dummy device for NAPI poll */ 1988 } reg_state:8; 1989 1990 bool dismantle; 1991 1992 enum { 1993 RTNL_LINK_INITIALIZED, 1994 RTNL_LINK_INITIALIZING, 1995 } rtnl_link_state:16; 1996 1997 bool needs_free_netdev; 1998 void (*priv_destructor)(struct net_device *dev); 1999 2000 #ifdef CONFIG_NETPOLL 2001 struct netpoll_info __rcu *npinfo; 2002 #endif 2003 2004 possible_net_t nd_net; 2005 2006 /* mid-layer private */ 2007 union { 2008 void *ml_priv; 2009 struct pcpu_lstats __percpu *lstats; 2010 struct pcpu_sw_netstats __percpu *tstats; 2011 struct pcpu_dstats __percpu *dstats; 2012 }; 2013 2014 #if IS_ENABLED(CONFIG_GARP) 2015 struct garp_port __rcu *garp_port; 2016 #endif 2017 #if IS_ENABLED(CONFIG_MRP) 2018 struct mrp_port __rcu *mrp_port; 2019 #endif 2020 2021 struct device dev; 2022 const struct attribute_group *sysfs_groups[4]; 2023 const struct attribute_group *sysfs_rx_queue_group; 2024 2025 const struct rtnl_link_ops *rtnl_link_ops; 2026 2027 /* for setting kernel sock attribute on TCP connection setup */ 2028 #define GSO_MAX_SIZE 65536 2029 unsigned int gso_max_size; 2030 #define GSO_MAX_SEGS 65535 2031 u16 gso_max_segs; 2032 2033 #ifdef CONFIG_DCB 2034 const struct dcbnl_rtnl_ops *dcbnl_ops; 2035 #endif 2036 s16 num_tc; 2037 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2038 u8 prio_tc_map[TC_BITMASK + 1]; 2039 2040 #if IS_ENABLED(CONFIG_FCOE) 2041 unsigned int fcoe_ddp_xid; 2042 #endif 2043 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2044 struct netprio_map __rcu *priomap; 2045 #endif 2046 struct phy_device *phydev; 2047 struct sfp_bus *sfp_bus; 2048 struct lock_class_key *qdisc_tx_busylock; 2049 struct lock_class_key *qdisc_running_key; 2050 bool proto_down; 2051 unsigned wol_enabled:1; 2052 }; 2053 #define to_net_dev(d) container_of(d, struct net_device, dev) 2054 2055 static inline bool netif_elide_gro(const struct net_device *dev) 2056 { 2057 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2058 return true; 2059 return false; 2060 } 2061 2062 #define NETDEV_ALIGN 32 2063 2064 static inline 2065 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2066 { 2067 return dev->prio_tc_map[prio & TC_BITMASK]; 2068 } 2069 2070 static inline 2071 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2072 { 2073 if (tc >= dev->num_tc) 2074 return -EINVAL; 2075 2076 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2077 return 0; 2078 } 2079 2080 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2081 void netdev_reset_tc(struct net_device *dev); 2082 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2083 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2084 2085 static inline 2086 int netdev_get_num_tc(struct net_device *dev) 2087 { 2088 return dev->num_tc; 2089 } 2090 2091 void netdev_unbind_sb_channel(struct net_device *dev, 2092 struct net_device *sb_dev); 2093 int netdev_bind_sb_channel_queue(struct net_device *dev, 2094 struct net_device *sb_dev, 2095 u8 tc, u16 count, u16 offset); 2096 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2097 static inline int netdev_get_sb_channel(struct net_device *dev) 2098 { 2099 return max_t(int, -dev->num_tc, 0); 2100 } 2101 2102 static inline 2103 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2104 unsigned int index) 2105 { 2106 return &dev->_tx[index]; 2107 } 2108 2109 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2110 const struct sk_buff *skb) 2111 { 2112 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2113 } 2114 2115 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2116 void (*f)(struct net_device *, 2117 struct netdev_queue *, 2118 void *), 2119 void *arg) 2120 { 2121 unsigned int i; 2122 2123 for (i = 0; i < dev->num_tx_queues; i++) 2124 f(dev, &dev->_tx[i], arg); 2125 } 2126 2127 #define netdev_lockdep_set_classes(dev) \ 2128 { \ 2129 static struct lock_class_key qdisc_tx_busylock_key; \ 2130 static struct lock_class_key qdisc_running_key; \ 2131 static struct lock_class_key qdisc_xmit_lock_key; \ 2132 static struct lock_class_key dev_addr_list_lock_key; \ 2133 unsigned int i; \ 2134 \ 2135 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2136 (dev)->qdisc_running_key = &qdisc_running_key; \ 2137 lockdep_set_class(&(dev)->addr_list_lock, \ 2138 &dev_addr_list_lock_key); \ 2139 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2140 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2141 &qdisc_xmit_lock_key); \ 2142 } 2143 2144 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2145 struct net_device *sb_dev); 2146 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2147 struct sk_buff *skb, 2148 struct net_device *sb_dev); 2149 2150 /* returns the headroom that the master device needs to take in account 2151 * when forwarding to this dev 2152 */ 2153 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2154 { 2155 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2156 } 2157 2158 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2159 { 2160 if (dev->netdev_ops->ndo_set_rx_headroom) 2161 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2162 } 2163 2164 /* set the device rx headroom to the dev's default */ 2165 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2166 { 2167 netdev_set_rx_headroom(dev, -1); 2168 } 2169 2170 /* 2171 * Net namespace inlines 2172 */ 2173 static inline 2174 struct net *dev_net(const struct net_device *dev) 2175 { 2176 return read_pnet(&dev->nd_net); 2177 } 2178 2179 static inline 2180 void dev_net_set(struct net_device *dev, struct net *net) 2181 { 2182 write_pnet(&dev->nd_net, net); 2183 } 2184 2185 /** 2186 * netdev_priv - access network device private data 2187 * @dev: network device 2188 * 2189 * Get network device private data 2190 */ 2191 static inline void *netdev_priv(const struct net_device *dev) 2192 { 2193 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2194 } 2195 2196 /* Set the sysfs physical device reference for the network logical device 2197 * if set prior to registration will cause a symlink during initialization. 2198 */ 2199 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2200 2201 /* Set the sysfs device type for the network logical device to allow 2202 * fine-grained identification of different network device types. For 2203 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2204 */ 2205 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2206 2207 /* Default NAPI poll() weight 2208 * Device drivers are strongly advised to not use bigger value 2209 */ 2210 #define NAPI_POLL_WEIGHT 64 2211 2212 /** 2213 * netif_napi_add - initialize a NAPI context 2214 * @dev: network device 2215 * @napi: NAPI context 2216 * @poll: polling function 2217 * @weight: default weight 2218 * 2219 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2220 * *any* of the other NAPI-related functions. 2221 */ 2222 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2223 int (*poll)(struct napi_struct *, int), int weight); 2224 2225 /** 2226 * netif_tx_napi_add - initialize a NAPI context 2227 * @dev: network device 2228 * @napi: NAPI context 2229 * @poll: polling function 2230 * @weight: default weight 2231 * 2232 * This variant of netif_napi_add() should be used from drivers using NAPI 2233 * to exclusively poll a TX queue. 2234 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2235 */ 2236 static inline void netif_tx_napi_add(struct net_device *dev, 2237 struct napi_struct *napi, 2238 int (*poll)(struct napi_struct *, int), 2239 int weight) 2240 { 2241 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2242 netif_napi_add(dev, napi, poll, weight); 2243 } 2244 2245 /** 2246 * netif_napi_del - remove a NAPI context 2247 * @napi: NAPI context 2248 * 2249 * netif_napi_del() removes a NAPI context from the network device NAPI list 2250 */ 2251 void netif_napi_del(struct napi_struct *napi); 2252 2253 struct napi_gro_cb { 2254 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2255 void *frag0; 2256 2257 /* Length of frag0. */ 2258 unsigned int frag0_len; 2259 2260 /* This indicates where we are processing relative to skb->data. */ 2261 int data_offset; 2262 2263 /* This is non-zero if the packet cannot be merged with the new skb. */ 2264 u16 flush; 2265 2266 /* Save the IP ID here and check when we get to the transport layer */ 2267 u16 flush_id; 2268 2269 /* Number of segments aggregated. */ 2270 u16 count; 2271 2272 /* Start offset for remote checksum offload */ 2273 u16 gro_remcsum_start; 2274 2275 /* jiffies when first packet was created/queued */ 2276 unsigned long age; 2277 2278 /* Used in ipv6_gro_receive() and foo-over-udp */ 2279 u16 proto; 2280 2281 /* This is non-zero if the packet may be of the same flow. */ 2282 u8 same_flow:1; 2283 2284 /* Used in tunnel GRO receive */ 2285 u8 encap_mark:1; 2286 2287 /* GRO checksum is valid */ 2288 u8 csum_valid:1; 2289 2290 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2291 u8 csum_cnt:3; 2292 2293 /* Free the skb? */ 2294 u8 free:2; 2295 #define NAPI_GRO_FREE 1 2296 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2297 2298 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2299 u8 is_ipv6:1; 2300 2301 /* Used in GRE, set in fou/gue_gro_receive */ 2302 u8 is_fou:1; 2303 2304 /* Used to determine if flush_id can be ignored */ 2305 u8 is_atomic:1; 2306 2307 /* Number of gro_receive callbacks this packet already went through */ 2308 u8 recursion_counter:4; 2309 2310 /* 1 bit hole */ 2311 2312 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2313 __wsum csum; 2314 2315 /* used in skb_gro_receive() slow path */ 2316 struct sk_buff *last; 2317 }; 2318 2319 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2320 2321 #define GRO_RECURSION_LIMIT 15 2322 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2323 { 2324 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2325 } 2326 2327 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2328 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2329 struct list_head *head, 2330 struct sk_buff *skb) 2331 { 2332 if (unlikely(gro_recursion_inc_test(skb))) { 2333 NAPI_GRO_CB(skb)->flush |= 1; 2334 return NULL; 2335 } 2336 2337 return cb(head, skb); 2338 } 2339 2340 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2341 struct sk_buff *); 2342 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2343 struct sock *sk, 2344 struct list_head *head, 2345 struct sk_buff *skb) 2346 { 2347 if (unlikely(gro_recursion_inc_test(skb))) { 2348 NAPI_GRO_CB(skb)->flush |= 1; 2349 return NULL; 2350 } 2351 2352 return cb(sk, head, skb); 2353 } 2354 2355 struct packet_type { 2356 __be16 type; /* This is really htons(ether_type). */ 2357 bool ignore_outgoing; 2358 struct net_device *dev; /* NULL is wildcarded here */ 2359 int (*func) (struct sk_buff *, 2360 struct net_device *, 2361 struct packet_type *, 2362 struct net_device *); 2363 void (*list_func) (struct list_head *, 2364 struct packet_type *, 2365 struct net_device *); 2366 bool (*id_match)(struct packet_type *ptype, 2367 struct sock *sk); 2368 void *af_packet_priv; 2369 struct list_head list; 2370 }; 2371 2372 struct offload_callbacks { 2373 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2374 netdev_features_t features); 2375 struct sk_buff *(*gro_receive)(struct list_head *head, 2376 struct sk_buff *skb); 2377 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2378 }; 2379 2380 struct packet_offload { 2381 __be16 type; /* This is really htons(ether_type). */ 2382 u16 priority; 2383 struct offload_callbacks callbacks; 2384 struct list_head list; 2385 }; 2386 2387 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2388 struct pcpu_sw_netstats { 2389 u64 rx_packets; 2390 u64 rx_bytes; 2391 u64 tx_packets; 2392 u64 tx_bytes; 2393 struct u64_stats_sync syncp; 2394 } __aligned(4 * sizeof(u64)); 2395 2396 struct pcpu_lstats { 2397 u64 packets; 2398 u64 bytes; 2399 struct u64_stats_sync syncp; 2400 } __aligned(2 * sizeof(u64)); 2401 2402 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2403 ({ \ 2404 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2405 if (pcpu_stats) { \ 2406 int __cpu; \ 2407 for_each_possible_cpu(__cpu) { \ 2408 typeof(type) *stat; \ 2409 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2410 u64_stats_init(&stat->syncp); \ 2411 } \ 2412 } \ 2413 pcpu_stats; \ 2414 }) 2415 2416 #define netdev_alloc_pcpu_stats(type) \ 2417 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2418 2419 enum netdev_lag_tx_type { 2420 NETDEV_LAG_TX_TYPE_UNKNOWN, 2421 NETDEV_LAG_TX_TYPE_RANDOM, 2422 NETDEV_LAG_TX_TYPE_BROADCAST, 2423 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2424 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2425 NETDEV_LAG_TX_TYPE_HASH, 2426 }; 2427 2428 enum netdev_lag_hash { 2429 NETDEV_LAG_HASH_NONE, 2430 NETDEV_LAG_HASH_L2, 2431 NETDEV_LAG_HASH_L34, 2432 NETDEV_LAG_HASH_L23, 2433 NETDEV_LAG_HASH_E23, 2434 NETDEV_LAG_HASH_E34, 2435 NETDEV_LAG_HASH_UNKNOWN, 2436 }; 2437 2438 struct netdev_lag_upper_info { 2439 enum netdev_lag_tx_type tx_type; 2440 enum netdev_lag_hash hash_type; 2441 }; 2442 2443 struct netdev_lag_lower_state_info { 2444 u8 link_up : 1, 2445 tx_enabled : 1; 2446 }; 2447 2448 #include <linux/notifier.h> 2449 2450 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2451 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2452 * adding new types. 2453 */ 2454 enum netdev_cmd { 2455 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2456 NETDEV_DOWN, 2457 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2458 detected a hardware crash and restarted 2459 - we can use this eg to kick tcp sessions 2460 once done */ 2461 NETDEV_CHANGE, /* Notify device state change */ 2462 NETDEV_REGISTER, 2463 NETDEV_UNREGISTER, 2464 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2465 NETDEV_CHANGEADDR, /* notify after the address change */ 2466 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2467 NETDEV_GOING_DOWN, 2468 NETDEV_CHANGENAME, 2469 NETDEV_FEAT_CHANGE, 2470 NETDEV_BONDING_FAILOVER, 2471 NETDEV_PRE_UP, 2472 NETDEV_PRE_TYPE_CHANGE, 2473 NETDEV_POST_TYPE_CHANGE, 2474 NETDEV_POST_INIT, 2475 NETDEV_RELEASE, 2476 NETDEV_NOTIFY_PEERS, 2477 NETDEV_JOIN, 2478 NETDEV_CHANGEUPPER, 2479 NETDEV_RESEND_IGMP, 2480 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2481 NETDEV_CHANGEINFODATA, 2482 NETDEV_BONDING_INFO, 2483 NETDEV_PRECHANGEUPPER, 2484 NETDEV_CHANGELOWERSTATE, 2485 NETDEV_UDP_TUNNEL_PUSH_INFO, 2486 NETDEV_UDP_TUNNEL_DROP_INFO, 2487 NETDEV_CHANGE_TX_QUEUE_LEN, 2488 NETDEV_CVLAN_FILTER_PUSH_INFO, 2489 NETDEV_CVLAN_FILTER_DROP_INFO, 2490 NETDEV_SVLAN_FILTER_PUSH_INFO, 2491 NETDEV_SVLAN_FILTER_DROP_INFO, 2492 }; 2493 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2494 2495 int register_netdevice_notifier(struct notifier_block *nb); 2496 int unregister_netdevice_notifier(struct notifier_block *nb); 2497 2498 struct netdev_notifier_info { 2499 struct net_device *dev; 2500 struct netlink_ext_ack *extack; 2501 }; 2502 2503 struct netdev_notifier_info_ext { 2504 struct netdev_notifier_info info; /* must be first */ 2505 union { 2506 u32 mtu; 2507 } ext; 2508 }; 2509 2510 struct netdev_notifier_change_info { 2511 struct netdev_notifier_info info; /* must be first */ 2512 unsigned int flags_changed; 2513 }; 2514 2515 struct netdev_notifier_changeupper_info { 2516 struct netdev_notifier_info info; /* must be first */ 2517 struct net_device *upper_dev; /* new upper dev */ 2518 bool master; /* is upper dev master */ 2519 bool linking; /* is the notification for link or unlink */ 2520 void *upper_info; /* upper dev info */ 2521 }; 2522 2523 struct netdev_notifier_changelowerstate_info { 2524 struct netdev_notifier_info info; /* must be first */ 2525 void *lower_state_info; /* is lower dev state */ 2526 }; 2527 2528 struct netdev_notifier_pre_changeaddr_info { 2529 struct netdev_notifier_info info; /* must be first */ 2530 const unsigned char *dev_addr; 2531 }; 2532 2533 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2534 struct net_device *dev) 2535 { 2536 info->dev = dev; 2537 info->extack = NULL; 2538 } 2539 2540 static inline struct net_device * 2541 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2542 { 2543 return info->dev; 2544 } 2545 2546 static inline struct netlink_ext_ack * 2547 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2548 { 2549 return info->extack; 2550 } 2551 2552 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2553 2554 2555 extern rwlock_t dev_base_lock; /* Device list lock */ 2556 2557 #define for_each_netdev(net, d) \ 2558 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2559 #define for_each_netdev_reverse(net, d) \ 2560 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2561 #define for_each_netdev_rcu(net, d) \ 2562 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2563 #define for_each_netdev_safe(net, d, n) \ 2564 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2565 #define for_each_netdev_continue(net, d) \ 2566 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2567 #define for_each_netdev_continue_rcu(net, d) \ 2568 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2569 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2570 for_each_netdev_rcu(&init_net, slave) \ 2571 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2572 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2573 2574 static inline struct net_device *next_net_device(struct net_device *dev) 2575 { 2576 struct list_head *lh; 2577 struct net *net; 2578 2579 net = dev_net(dev); 2580 lh = dev->dev_list.next; 2581 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2582 } 2583 2584 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2585 { 2586 struct list_head *lh; 2587 struct net *net; 2588 2589 net = dev_net(dev); 2590 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2591 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2592 } 2593 2594 static inline struct net_device *first_net_device(struct net *net) 2595 { 2596 return list_empty(&net->dev_base_head) ? NULL : 2597 net_device_entry(net->dev_base_head.next); 2598 } 2599 2600 static inline struct net_device *first_net_device_rcu(struct net *net) 2601 { 2602 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2603 2604 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2605 } 2606 2607 int netdev_boot_setup_check(struct net_device *dev); 2608 unsigned long netdev_boot_base(const char *prefix, int unit); 2609 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2610 const char *hwaddr); 2611 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2612 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2613 void dev_add_pack(struct packet_type *pt); 2614 void dev_remove_pack(struct packet_type *pt); 2615 void __dev_remove_pack(struct packet_type *pt); 2616 void dev_add_offload(struct packet_offload *po); 2617 void dev_remove_offload(struct packet_offload *po); 2618 2619 int dev_get_iflink(const struct net_device *dev); 2620 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2621 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2622 unsigned short mask); 2623 struct net_device *dev_get_by_name(struct net *net, const char *name); 2624 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2625 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2626 int dev_alloc_name(struct net_device *dev, const char *name); 2627 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2628 void dev_close(struct net_device *dev); 2629 void dev_close_many(struct list_head *head, bool unlink); 2630 void dev_disable_lro(struct net_device *dev); 2631 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2632 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2633 struct net_device *sb_dev); 2634 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2635 struct net_device *sb_dev); 2636 int dev_queue_xmit(struct sk_buff *skb); 2637 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2638 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2639 int register_netdevice(struct net_device *dev); 2640 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2641 void unregister_netdevice_many(struct list_head *head); 2642 static inline void unregister_netdevice(struct net_device *dev) 2643 { 2644 unregister_netdevice_queue(dev, NULL); 2645 } 2646 2647 int netdev_refcnt_read(const struct net_device *dev); 2648 void free_netdev(struct net_device *dev); 2649 void netdev_freemem(struct net_device *dev); 2650 void synchronize_net(void); 2651 int init_dummy_netdev(struct net_device *dev); 2652 2653 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2654 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2655 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2656 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2657 int netdev_get_name(struct net *net, char *name, int ifindex); 2658 int dev_restart(struct net_device *dev); 2659 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2660 2661 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2662 { 2663 return NAPI_GRO_CB(skb)->data_offset; 2664 } 2665 2666 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2667 { 2668 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2669 } 2670 2671 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2672 { 2673 NAPI_GRO_CB(skb)->data_offset += len; 2674 } 2675 2676 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2677 unsigned int offset) 2678 { 2679 return NAPI_GRO_CB(skb)->frag0 + offset; 2680 } 2681 2682 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2683 { 2684 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2685 } 2686 2687 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2688 { 2689 NAPI_GRO_CB(skb)->frag0 = NULL; 2690 NAPI_GRO_CB(skb)->frag0_len = 0; 2691 } 2692 2693 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2694 unsigned int offset) 2695 { 2696 if (!pskb_may_pull(skb, hlen)) 2697 return NULL; 2698 2699 skb_gro_frag0_invalidate(skb); 2700 return skb->data + offset; 2701 } 2702 2703 static inline void *skb_gro_network_header(struct sk_buff *skb) 2704 { 2705 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2706 skb_network_offset(skb); 2707 } 2708 2709 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2710 const void *start, unsigned int len) 2711 { 2712 if (NAPI_GRO_CB(skb)->csum_valid) 2713 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2714 csum_partial(start, len, 0)); 2715 } 2716 2717 /* GRO checksum functions. These are logical equivalents of the normal 2718 * checksum functions (in skbuff.h) except that they operate on the GRO 2719 * offsets and fields in sk_buff. 2720 */ 2721 2722 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2723 2724 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2725 { 2726 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2727 } 2728 2729 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2730 bool zero_okay, 2731 __sum16 check) 2732 { 2733 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2734 skb_checksum_start_offset(skb) < 2735 skb_gro_offset(skb)) && 2736 !skb_at_gro_remcsum_start(skb) && 2737 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2738 (!zero_okay || check)); 2739 } 2740 2741 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2742 __wsum psum) 2743 { 2744 if (NAPI_GRO_CB(skb)->csum_valid && 2745 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2746 return 0; 2747 2748 NAPI_GRO_CB(skb)->csum = psum; 2749 2750 return __skb_gro_checksum_complete(skb); 2751 } 2752 2753 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2754 { 2755 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2756 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2757 NAPI_GRO_CB(skb)->csum_cnt--; 2758 } else { 2759 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2760 * verified a new top level checksum or an encapsulated one 2761 * during GRO. This saves work if we fallback to normal path. 2762 */ 2763 __skb_incr_checksum_unnecessary(skb); 2764 } 2765 } 2766 2767 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2768 compute_pseudo) \ 2769 ({ \ 2770 __sum16 __ret = 0; \ 2771 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2772 __ret = __skb_gro_checksum_validate_complete(skb, \ 2773 compute_pseudo(skb, proto)); \ 2774 if (!__ret) \ 2775 skb_gro_incr_csum_unnecessary(skb); \ 2776 __ret; \ 2777 }) 2778 2779 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2780 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2781 2782 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2783 compute_pseudo) \ 2784 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2785 2786 #define skb_gro_checksum_simple_validate(skb) \ 2787 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2788 2789 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2790 { 2791 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2792 !NAPI_GRO_CB(skb)->csum_valid); 2793 } 2794 2795 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2796 __sum16 check, __wsum pseudo) 2797 { 2798 NAPI_GRO_CB(skb)->csum = ~pseudo; 2799 NAPI_GRO_CB(skb)->csum_valid = 1; 2800 } 2801 2802 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2803 do { \ 2804 if (__skb_gro_checksum_convert_check(skb)) \ 2805 __skb_gro_checksum_convert(skb, check, \ 2806 compute_pseudo(skb, proto)); \ 2807 } while (0) 2808 2809 struct gro_remcsum { 2810 int offset; 2811 __wsum delta; 2812 }; 2813 2814 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2815 { 2816 grc->offset = 0; 2817 grc->delta = 0; 2818 } 2819 2820 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2821 unsigned int off, size_t hdrlen, 2822 int start, int offset, 2823 struct gro_remcsum *grc, 2824 bool nopartial) 2825 { 2826 __wsum delta; 2827 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2828 2829 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2830 2831 if (!nopartial) { 2832 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2833 return ptr; 2834 } 2835 2836 ptr = skb_gro_header_fast(skb, off); 2837 if (skb_gro_header_hard(skb, off + plen)) { 2838 ptr = skb_gro_header_slow(skb, off + plen, off); 2839 if (!ptr) 2840 return NULL; 2841 } 2842 2843 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2844 start, offset); 2845 2846 /* Adjust skb->csum since we changed the packet */ 2847 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2848 2849 grc->offset = off + hdrlen + offset; 2850 grc->delta = delta; 2851 2852 return ptr; 2853 } 2854 2855 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2856 struct gro_remcsum *grc) 2857 { 2858 void *ptr; 2859 size_t plen = grc->offset + sizeof(u16); 2860 2861 if (!grc->delta) 2862 return; 2863 2864 ptr = skb_gro_header_fast(skb, grc->offset); 2865 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2866 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2867 if (!ptr) 2868 return; 2869 } 2870 2871 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2872 } 2873 2874 #ifdef CONFIG_XFRM_OFFLOAD 2875 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2876 { 2877 if (PTR_ERR(pp) != -EINPROGRESS) 2878 NAPI_GRO_CB(skb)->flush |= flush; 2879 } 2880 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2881 struct sk_buff *pp, 2882 int flush, 2883 struct gro_remcsum *grc) 2884 { 2885 if (PTR_ERR(pp) != -EINPROGRESS) { 2886 NAPI_GRO_CB(skb)->flush |= flush; 2887 skb_gro_remcsum_cleanup(skb, grc); 2888 skb->remcsum_offload = 0; 2889 } 2890 } 2891 #else 2892 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2893 { 2894 NAPI_GRO_CB(skb)->flush |= flush; 2895 } 2896 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2897 struct sk_buff *pp, 2898 int flush, 2899 struct gro_remcsum *grc) 2900 { 2901 NAPI_GRO_CB(skb)->flush |= flush; 2902 skb_gro_remcsum_cleanup(skb, grc); 2903 skb->remcsum_offload = 0; 2904 } 2905 #endif 2906 2907 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2908 unsigned short type, 2909 const void *daddr, const void *saddr, 2910 unsigned int len) 2911 { 2912 if (!dev->header_ops || !dev->header_ops->create) 2913 return 0; 2914 2915 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2916 } 2917 2918 static inline int dev_parse_header(const struct sk_buff *skb, 2919 unsigned char *haddr) 2920 { 2921 const struct net_device *dev = skb->dev; 2922 2923 if (!dev->header_ops || !dev->header_ops->parse) 2924 return 0; 2925 return dev->header_ops->parse(skb, haddr); 2926 } 2927 2928 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 2929 { 2930 const struct net_device *dev = skb->dev; 2931 2932 if (!dev->header_ops || !dev->header_ops->parse_protocol) 2933 return 0; 2934 return dev->header_ops->parse_protocol(skb); 2935 } 2936 2937 /* ll_header must have at least hard_header_len allocated */ 2938 static inline bool dev_validate_header(const struct net_device *dev, 2939 char *ll_header, int len) 2940 { 2941 if (likely(len >= dev->hard_header_len)) 2942 return true; 2943 if (len < dev->min_header_len) 2944 return false; 2945 2946 if (capable(CAP_SYS_RAWIO)) { 2947 memset(ll_header + len, 0, dev->hard_header_len - len); 2948 return true; 2949 } 2950 2951 if (dev->header_ops && dev->header_ops->validate) 2952 return dev->header_ops->validate(ll_header, len); 2953 2954 return false; 2955 } 2956 2957 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 2958 int len, int size); 2959 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2960 static inline int unregister_gifconf(unsigned int family) 2961 { 2962 return register_gifconf(family, NULL); 2963 } 2964 2965 #ifdef CONFIG_NET_FLOW_LIMIT 2966 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2967 struct sd_flow_limit { 2968 u64 count; 2969 unsigned int num_buckets; 2970 unsigned int history_head; 2971 u16 history[FLOW_LIMIT_HISTORY]; 2972 u8 buckets[]; 2973 }; 2974 2975 extern int netdev_flow_limit_table_len; 2976 #endif /* CONFIG_NET_FLOW_LIMIT */ 2977 2978 /* 2979 * Incoming packets are placed on per-CPU queues 2980 */ 2981 struct softnet_data { 2982 struct list_head poll_list; 2983 struct sk_buff_head process_queue; 2984 2985 /* stats */ 2986 unsigned int processed; 2987 unsigned int time_squeeze; 2988 unsigned int received_rps; 2989 #ifdef CONFIG_RPS 2990 struct softnet_data *rps_ipi_list; 2991 #endif 2992 #ifdef CONFIG_NET_FLOW_LIMIT 2993 struct sd_flow_limit __rcu *flow_limit; 2994 #endif 2995 struct Qdisc *output_queue; 2996 struct Qdisc **output_queue_tailp; 2997 struct sk_buff *completion_queue; 2998 #ifdef CONFIG_XFRM_OFFLOAD 2999 struct sk_buff_head xfrm_backlog; 3000 #endif 3001 /* written and read only by owning cpu: */ 3002 struct { 3003 u16 recursion; 3004 u8 more; 3005 } xmit; 3006 #ifdef CONFIG_RPS 3007 /* input_queue_head should be written by cpu owning this struct, 3008 * and only read by other cpus. Worth using a cache line. 3009 */ 3010 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3011 3012 /* Elements below can be accessed between CPUs for RPS/RFS */ 3013 call_single_data_t csd ____cacheline_aligned_in_smp; 3014 struct softnet_data *rps_ipi_next; 3015 unsigned int cpu; 3016 unsigned int input_queue_tail; 3017 #endif 3018 unsigned int dropped; 3019 struct sk_buff_head input_pkt_queue; 3020 struct napi_struct backlog; 3021 3022 }; 3023 3024 static inline void input_queue_head_incr(struct softnet_data *sd) 3025 { 3026 #ifdef CONFIG_RPS 3027 sd->input_queue_head++; 3028 #endif 3029 } 3030 3031 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3032 unsigned int *qtail) 3033 { 3034 #ifdef CONFIG_RPS 3035 *qtail = ++sd->input_queue_tail; 3036 #endif 3037 } 3038 3039 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3040 3041 static inline int dev_recursion_level(void) 3042 { 3043 return this_cpu_read(softnet_data.xmit.recursion); 3044 } 3045 3046 #define XMIT_RECURSION_LIMIT 10 3047 static inline bool dev_xmit_recursion(void) 3048 { 3049 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3050 XMIT_RECURSION_LIMIT); 3051 } 3052 3053 static inline void dev_xmit_recursion_inc(void) 3054 { 3055 __this_cpu_inc(softnet_data.xmit.recursion); 3056 } 3057 3058 static inline void dev_xmit_recursion_dec(void) 3059 { 3060 __this_cpu_dec(softnet_data.xmit.recursion); 3061 } 3062 3063 void __netif_schedule(struct Qdisc *q); 3064 void netif_schedule_queue(struct netdev_queue *txq); 3065 3066 static inline void netif_tx_schedule_all(struct net_device *dev) 3067 { 3068 unsigned int i; 3069 3070 for (i = 0; i < dev->num_tx_queues; i++) 3071 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3072 } 3073 3074 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3075 { 3076 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3077 } 3078 3079 /** 3080 * netif_start_queue - allow transmit 3081 * @dev: network device 3082 * 3083 * Allow upper layers to call the device hard_start_xmit routine. 3084 */ 3085 static inline void netif_start_queue(struct net_device *dev) 3086 { 3087 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3088 } 3089 3090 static inline void netif_tx_start_all_queues(struct net_device *dev) 3091 { 3092 unsigned int i; 3093 3094 for (i = 0; i < dev->num_tx_queues; i++) { 3095 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3096 netif_tx_start_queue(txq); 3097 } 3098 } 3099 3100 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3101 3102 /** 3103 * netif_wake_queue - restart transmit 3104 * @dev: network device 3105 * 3106 * Allow upper layers to call the device hard_start_xmit routine. 3107 * Used for flow control when transmit resources are available. 3108 */ 3109 static inline void netif_wake_queue(struct net_device *dev) 3110 { 3111 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3112 } 3113 3114 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3115 { 3116 unsigned int i; 3117 3118 for (i = 0; i < dev->num_tx_queues; i++) { 3119 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3120 netif_tx_wake_queue(txq); 3121 } 3122 } 3123 3124 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3125 { 3126 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3127 } 3128 3129 /** 3130 * netif_stop_queue - stop transmitted packets 3131 * @dev: network device 3132 * 3133 * Stop upper layers calling the device hard_start_xmit routine. 3134 * Used for flow control when transmit resources are unavailable. 3135 */ 3136 static inline void netif_stop_queue(struct net_device *dev) 3137 { 3138 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3139 } 3140 3141 void netif_tx_stop_all_queues(struct net_device *dev); 3142 3143 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3144 { 3145 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3146 } 3147 3148 /** 3149 * netif_queue_stopped - test if transmit queue is flowblocked 3150 * @dev: network device 3151 * 3152 * Test if transmit queue on device is currently unable to send. 3153 */ 3154 static inline bool netif_queue_stopped(const struct net_device *dev) 3155 { 3156 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3157 } 3158 3159 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3160 { 3161 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3162 } 3163 3164 static inline bool 3165 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3166 { 3167 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3168 } 3169 3170 static inline bool 3171 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3172 { 3173 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3174 } 3175 3176 /** 3177 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3178 * @dev_queue: pointer to transmit queue 3179 * 3180 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3181 * to give appropriate hint to the CPU. 3182 */ 3183 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3184 { 3185 #ifdef CONFIG_BQL 3186 prefetchw(&dev_queue->dql.num_queued); 3187 #endif 3188 } 3189 3190 /** 3191 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3192 * @dev_queue: pointer to transmit queue 3193 * 3194 * BQL enabled drivers might use this helper in their TX completion path, 3195 * to give appropriate hint to the CPU. 3196 */ 3197 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3198 { 3199 #ifdef CONFIG_BQL 3200 prefetchw(&dev_queue->dql.limit); 3201 #endif 3202 } 3203 3204 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3205 unsigned int bytes) 3206 { 3207 #ifdef CONFIG_BQL 3208 dql_queued(&dev_queue->dql, bytes); 3209 3210 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3211 return; 3212 3213 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3214 3215 /* 3216 * The XOFF flag must be set before checking the dql_avail below, 3217 * because in netdev_tx_completed_queue we update the dql_completed 3218 * before checking the XOFF flag. 3219 */ 3220 smp_mb(); 3221 3222 /* check again in case another CPU has just made room avail */ 3223 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3224 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3225 #endif 3226 } 3227 3228 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3229 * that they should not test BQL status themselves. 3230 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3231 * skb of a batch. 3232 * Returns true if the doorbell must be used to kick the NIC. 3233 */ 3234 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3235 unsigned int bytes, 3236 bool xmit_more) 3237 { 3238 if (xmit_more) { 3239 #ifdef CONFIG_BQL 3240 dql_queued(&dev_queue->dql, bytes); 3241 #endif 3242 return netif_tx_queue_stopped(dev_queue); 3243 } 3244 netdev_tx_sent_queue(dev_queue, bytes); 3245 return true; 3246 } 3247 3248 /** 3249 * netdev_sent_queue - report the number of bytes queued to hardware 3250 * @dev: network device 3251 * @bytes: number of bytes queued to the hardware device queue 3252 * 3253 * Report the number of bytes queued for sending/completion to the network 3254 * device hardware queue. @bytes should be a good approximation and should 3255 * exactly match netdev_completed_queue() @bytes 3256 */ 3257 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3258 { 3259 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3260 } 3261 3262 static inline bool __netdev_sent_queue(struct net_device *dev, 3263 unsigned int bytes, 3264 bool xmit_more) 3265 { 3266 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3267 xmit_more); 3268 } 3269 3270 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3271 unsigned int pkts, unsigned int bytes) 3272 { 3273 #ifdef CONFIG_BQL 3274 if (unlikely(!bytes)) 3275 return; 3276 3277 dql_completed(&dev_queue->dql, bytes); 3278 3279 /* 3280 * Without the memory barrier there is a small possiblity that 3281 * netdev_tx_sent_queue will miss the update and cause the queue to 3282 * be stopped forever 3283 */ 3284 smp_mb(); 3285 3286 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3287 return; 3288 3289 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3290 netif_schedule_queue(dev_queue); 3291 #endif 3292 } 3293 3294 /** 3295 * netdev_completed_queue - report bytes and packets completed by device 3296 * @dev: network device 3297 * @pkts: actual number of packets sent over the medium 3298 * @bytes: actual number of bytes sent over the medium 3299 * 3300 * Report the number of bytes and packets transmitted by the network device 3301 * hardware queue over the physical medium, @bytes must exactly match the 3302 * @bytes amount passed to netdev_sent_queue() 3303 */ 3304 static inline void netdev_completed_queue(struct net_device *dev, 3305 unsigned int pkts, unsigned int bytes) 3306 { 3307 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3308 } 3309 3310 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3311 { 3312 #ifdef CONFIG_BQL 3313 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3314 dql_reset(&q->dql); 3315 #endif 3316 } 3317 3318 /** 3319 * netdev_reset_queue - reset the packets and bytes count of a network device 3320 * @dev_queue: network device 3321 * 3322 * Reset the bytes and packet count of a network device and clear the 3323 * software flow control OFF bit for this network device 3324 */ 3325 static inline void netdev_reset_queue(struct net_device *dev_queue) 3326 { 3327 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3328 } 3329 3330 /** 3331 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3332 * @dev: network device 3333 * @queue_index: given tx queue index 3334 * 3335 * Returns 0 if given tx queue index >= number of device tx queues, 3336 * otherwise returns the originally passed tx queue index. 3337 */ 3338 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3339 { 3340 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3341 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3342 dev->name, queue_index, 3343 dev->real_num_tx_queues); 3344 return 0; 3345 } 3346 3347 return queue_index; 3348 } 3349 3350 /** 3351 * netif_running - test if up 3352 * @dev: network device 3353 * 3354 * Test if the device has been brought up. 3355 */ 3356 static inline bool netif_running(const struct net_device *dev) 3357 { 3358 return test_bit(__LINK_STATE_START, &dev->state); 3359 } 3360 3361 /* 3362 * Routines to manage the subqueues on a device. We only need start, 3363 * stop, and a check if it's stopped. All other device management is 3364 * done at the overall netdevice level. 3365 * Also test the device if we're multiqueue. 3366 */ 3367 3368 /** 3369 * netif_start_subqueue - allow sending packets on subqueue 3370 * @dev: network device 3371 * @queue_index: sub queue index 3372 * 3373 * Start individual transmit queue of a device with multiple transmit queues. 3374 */ 3375 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3376 { 3377 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3378 3379 netif_tx_start_queue(txq); 3380 } 3381 3382 /** 3383 * netif_stop_subqueue - stop sending packets on subqueue 3384 * @dev: network device 3385 * @queue_index: sub queue index 3386 * 3387 * Stop individual transmit queue of a device with multiple transmit queues. 3388 */ 3389 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3390 { 3391 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3392 netif_tx_stop_queue(txq); 3393 } 3394 3395 /** 3396 * netif_subqueue_stopped - test status of subqueue 3397 * @dev: network device 3398 * @queue_index: sub queue index 3399 * 3400 * Check individual transmit queue of a device with multiple transmit queues. 3401 */ 3402 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3403 u16 queue_index) 3404 { 3405 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3406 3407 return netif_tx_queue_stopped(txq); 3408 } 3409 3410 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3411 struct sk_buff *skb) 3412 { 3413 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3414 } 3415 3416 /** 3417 * netif_wake_subqueue - allow sending packets on subqueue 3418 * @dev: network device 3419 * @queue_index: sub queue index 3420 * 3421 * Resume individual transmit queue of a device with multiple transmit queues. 3422 */ 3423 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3424 { 3425 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3426 3427 netif_tx_wake_queue(txq); 3428 } 3429 3430 #ifdef CONFIG_XPS 3431 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3432 u16 index); 3433 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3434 u16 index, bool is_rxqs_map); 3435 3436 /** 3437 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3438 * @j: CPU/Rx queue index 3439 * @mask: bitmask of all cpus/rx queues 3440 * @nr_bits: number of bits in the bitmask 3441 * 3442 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3443 */ 3444 static inline bool netif_attr_test_mask(unsigned long j, 3445 const unsigned long *mask, 3446 unsigned int nr_bits) 3447 { 3448 cpu_max_bits_warn(j, nr_bits); 3449 return test_bit(j, mask); 3450 } 3451 3452 /** 3453 * netif_attr_test_online - Test for online CPU/Rx queue 3454 * @j: CPU/Rx queue index 3455 * @online_mask: bitmask for CPUs/Rx queues that are online 3456 * @nr_bits: number of bits in the bitmask 3457 * 3458 * Returns true if a CPU/Rx queue is online. 3459 */ 3460 static inline bool netif_attr_test_online(unsigned long j, 3461 const unsigned long *online_mask, 3462 unsigned int nr_bits) 3463 { 3464 cpu_max_bits_warn(j, nr_bits); 3465 3466 if (online_mask) 3467 return test_bit(j, online_mask); 3468 3469 return (j < nr_bits); 3470 } 3471 3472 /** 3473 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3474 * @n: CPU/Rx queue index 3475 * @srcp: the cpumask/Rx queue mask pointer 3476 * @nr_bits: number of bits in the bitmask 3477 * 3478 * Returns >= nr_bits if no further CPUs/Rx queues set. 3479 */ 3480 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3481 unsigned int nr_bits) 3482 { 3483 /* -1 is a legal arg here. */ 3484 if (n != -1) 3485 cpu_max_bits_warn(n, nr_bits); 3486 3487 if (srcp) 3488 return find_next_bit(srcp, nr_bits, n + 1); 3489 3490 return n + 1; 3491 } 3492 3493 /** 3494 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p 3495 * @n: CPU/Rx queue index 3496 * @src1p: the first CPUs/Rx queues mask pointer 3497 * @src2p: the second CPUs/Rx queues mask pointer 3498 * @nr_bits: number of bits in the bitmask 3499 * 3500 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3501 */ 3502 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3503 const unsigned long *src2p, 3504 unsigned int nr_bits) 3505 { 3506 /* -1 is a legal arg here. */ 3507 if (n != -1) 3508 cpu_max_bits_warn(n, nr_bits); 3509 3510 if (src1p && src2p) 3511 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3512 else if (src1p) 3513 return find_next_bit(src1p, nr_bits, n + 1); 3514 else if (src2p) 3515 return find_next_bit(src2p, nr_bits, n + 1); 3516 3517 return n + 1; 3518 } 3519 #else 3520 static inline int netif_set_xps_queue(struct net_device *dev, 3521 const struct cpumask *mask, 3522 u16 index) 3523 { 3524 return 0; 3525 } 3526 3527 static inline int __netif_set_xps_queue(struct net_device *dev, 3528 const unsigned long *mask, 3529 u16 index, bool is_rxqs_map) 3530 { 3531 return 0; 3532 } 3533 #endif 3534 3535 /** 3536 * netif_is_multiqueue - test if device has multiple transmit queues 3537 * @dev: network device 3538 * 3539 * Check if device has multiple transmit queues 3540 */ 3541 static inline bool netif_is_multiqueue(const struct net_device *dev) 3542 { 3543 return dev->num_tx_queues > 1; 3544 } 3545 3546 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3547 3548 #ifdef CONFIG_SYSFS 3549 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3550 #else 3551 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3552 unsigned int rxqs) 3553 { 3554 dev->real_num_rx_queues = rxqs; 3555 return 0; 3556 } 3557 #endif 3558 3559 static inline struct netdev_rx_queue * 3560 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3561 { 3562 return dev->_rx + rxq; 3563 } 3564 3565 #ifdef CONFIG_SYSFS 3566 static inline unsigned int get_netdev_rx_queue_index( 3567 struct netdev_rx_queue *queue) 3568 { 3569 struct net_device *dev = queue->dev; 3570 int index = queue - dev->_rx; 3571 3572 BUG_ON(index >= dev->num_rx_queues); 3573 return index; 3574 } 3575 #endif 3576 3577 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3578 int netif_get_num_default_rss_queues(void); 3579 3580 enum skb_free_reason { 3581 SKB_REASON_CONSUMED, 3582 SKB_REASON_DROPPED, 3583 }; 3584 3585 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3586 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3587 3588 /* 3589 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3590 * interrupt context or with hardware interrupts being disabled. 3591 * (in_irq() || irqs_disabled()) 3592 * 3593 * We provide four helpers that can be used in following contexts : 3594 * 3595 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3596 * replacing kfree_skb(skb) 3597 * 3598 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3599 * Typically used in place of consume_skb(skb) in TX completion path 3600 * 3601 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3602 * replacing kfree_skb(skb) 3603 * 3604 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3605 * and consumed a packet. Used in place of consume_skb(skb) 3606 */ 3607 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3608 { 3609 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3610 } 3611 3612 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3613 { 3614 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3615 } 3616 3617 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3618 { 3619 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3620 } 3621 3622 static inline void dev_consume_skb_any(struct sk_buff *skb) 3623 { 3624 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3625 } 3626 3627 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3628 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3629 int netif_rx(struct sk_buff *skb); 3630 int netif_rx_ni(struct sk_buff *skb); 3631 int netif_receive_skb(struct sk_buff *skb); 3632 int netif_receive_skb_core(struct sk_buff *skb); 3633 void netif_receive_skb_list(struct list_head *head); 3634 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3635 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3636 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3637 gro_result_t napi_gro_frags(struct napi_struct *napi); 3638 struct packet_offload *gro_find_receive_by_type(__be16 type); 3639 struct packet_offload *gro_find_complete_by_type(__be16 type); 3640 3641 static inline void napi_free_frags(struct napi_struct *napi) 3642 { 3643 kfree_skb(napi->skb); 3644 napi->skb = NULL; 3645 } 3646 3647 bool netdev_is_rx_handler_busy(struct net_device *dev); 3648 int netdev_rx_handler_register(struct net_device *dev, 3649 rx_handler_func_t *rx_handler, 3650 void *rx_handler_data); 3651 void netdev_rx_handler_unregister(struct net_device *dev); 3652 3653 bool dev_valid_name(const char *name); 3654 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3655 bool *need_copyout); 3656 int dev_ifconf(struct net *net, struct ifconf *, int); 3657 int dev_ethtool(struct net *net, struct ifreq *); 3658 unsigned int dev_get_flags(const struct net_device *); 3659 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3660 struct netlink_ext_ack *extack); 3661 int dev_change_flags(struct net_device *dev, unsigned int flags, 3662 struct netlink_ext_ack *extack); 3663 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3664 unsigned int gchanges); 3665 int dev_change_name(struct net_device *, const char *); 3666 int dev_set_alias(struct net_device *, const char *, size_t); 3667 int dev_get_alias(const struct net_device *, char *, size_t); 3668 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3669 int __dev_set_mtu(struct net_device *, int); 3670 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3671 struct netlink_ext_ack *extack); 3672 int dev_set_mtu(struct net_device *, int); 3673 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3674 void dev_set_group(struct net_device *, int); 3675 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3676 struct netlink_ext_ack *extack); 3677 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3678 struct netlink_ext_ack *extack); 3679 int dev_change_carrier(struct net_device *, bool new_carrier); 3680 int dev_get_phys_port_id(struct net_device *dev, 3681 struct netdev_phys_item_id *ppid); 3682 int dev_get_phys_port_name(struct net_device *dev, 3683 char *name, size_t len); 3684 int dev_get_port_parent_id(struct net_device *dev, 3685 struct netdev_phys_item_id *ppid, bool recurse); 3686 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3687 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3688 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3689 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3690 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3691 struct netdev_queue *txq, int *ret); 3692 3693 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3694 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3695 int fd, u32 flags); 3696 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3697 enum bpf_netdev_command cmd); 3698 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3699 3700 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3701 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3702 bool is_skb_forwardable(const struct net_device *dev, 3703 const struct sk_buff *skb); 3704 3705 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3706 struct sk_buff *skb) 3707 { 3708 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3709 unlikely(!is_skb_forwardable(dev, skb))) { 3710 atomic_long_inc(&dev->rx_dropped); 3711 kfree_skb(skb); 3712 return NET_RX_DROP; 3713 } 3714 3715 skb_scrub_packet(skb, true); 3716 skb->priority = 0; 3717 return 0; 3718 } 3719 3720 bool dev_nit_active(struct net_device *dev); 3721 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3722 3723 extern int netdev_budget; 3724 extern unsigned int netdev_budget_usecs; 3725 3726 /* Called by rtnetlink.c:rtnl_unlock() */ 3727 void netdev_run_todo(void); 3728 3729 /** 3730 * dev_put - release reference to device 3731 * @dev: network device 3732 * 3733 * Release reference to device to allow it to be freed. 3734 */ 3735 static inline void dev_put(struct net_device *dev) 3736 { 3737 this_cpu_dec(*dev->pcpu_refcnt); 3738 } 3739 3740 /** 3741 * dev_hold - get reference to device 3742 * @dev: network device 3743 * 3744 * Hold reference to device to keep it from being freed. 3745 */ 3746 static inline void dev_hold(struct net_device *dev) 3747 { 3748 this_cpu_inc(*dev->pcpu_refcnt); 3749 } 3750 3751 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3752 * and _off may be called from IRQ context, but it is caller 3753 * who is responsible for serialization of these calls. 3754 * 3755 * The name carrier is inappropriate, these functions should really be 3756 * called netif_lowerlayer_*() because they represent the state of any 3757 * kind of lower layer not just hardware media. 3758 */ 3759 3760 void linkwatch_init_dev(struct net_device *dev); 3761 void linkwatch_fire_event(struct net_device *dev); 3762 void linkwatch_forget_dev(struct net_device *dev); 3763 3764 /** 3765 * netif_carrier_ok - test if carrier present 3766 * @dev: network device 3767 * 3768 * Check if carrier is present on device 3769 */ 3770 static inline bool netif_carrier_ok(const struct net_device *dev) 3771 { 3772 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3773 } 3774 3775 unsigned long dev_trans_start(struct net_device *dev); 3776 3777 void __netdev_watchdog_up(struct net_device *dev); 3778 3779 void netif_carrier_on(struct net_device *dev); 3780 3781 void netif_carrier_off(struct net_device *dev); 3782 3783 /** 3784 * netif_dormant_on - mark device as dormant. 3785 * @dev: network device 3786 * 3787 * Mark device as dormant (as per RFC2863). 3788 * 3789 * The dormant state indicates that the relevant interface is not 3790 * actually in a condition to pass packets (i.e., it is not 'up') but is 3791 * in a "pending" state, waiting for some external event. For "on- 3792 * demand" interfaces, this new state identifies the situation where the 3793 * interface is waiting for events to place it in the up state. 3794 */ 3795 static inline void netif_dormant_on(struct net_device *dev) 3796 { 3797 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3798 linkwatch_fire_event(dev); 3799 } 3800 3801 /** 3802 * netif_dormant_off - set device as not dormant. 3803 * @dev: network device 3804 * 3805 * Device is not in dormant state. 3806 */ 3807 static inline void netif_dormant_off(struct net_device *dev) 3808 { 3809 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3810 linkwatch_fire_event(dev); 3811 } 3812 3813 /** 3814 * netif_dormant - test if device is dormant 3815 * @dev: network device 3816 * 3817 * Check if device is dormant. 3818 */ 3819 static inline bool netif_dormant(const struct net_device *dev) 3820 { 3821 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3822 } 3823 3824 3825 /** 3826 * netif_oper_up - test if device is operational 3827 * @dev: network device 3828 * 3829 * Check if carrier is operational 3830 */ 3831 static inline bool netif_oper_up(const struct net_device *dev) 3832 { 3833 return (dev->operstate == IF_OPER_UP || 3834 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3835 } 3836 3837 /** 3838 * netif_device_present - is device available or removed 3839 * @dev: network device 3840 * 3841 * Check if device has not been removed from system. 3842 */ 3843 static inline bool netif_device_present(struct net_device *dev) 3844 { 3845 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3846 } 3847 3848 void netif_device_detach(struct net_device *dev); 3849 3850 void netif_device_attach(struct net_device *dev); 3851 3852 /* 3853 * Network interface message level settings 3854 */ 3855 3856 enum { 3857 NETIF_MSG_DRV = 0x0001, 3858 NETIF_MSG_PROBE = 0x0002, 3859 NETIF_MSG_LINK = 0x0004, 3860 NETIF_MSG_TIMER = 0x0008, 3861 NETIF_MSG_IFDOWN = 0x0010, 3862 NETIF_MSG_IFUP = 0x0020, 3863 NETIF_MSG_RX_ERR = 0x0040, 3864 NETIF_MSG_TX_ERR = 0x0080, 3865 NETIF_MSG_TX_QUEUED = 0x0100, 3866 NETIF_MSG_INTR = 0x0200, 3867 NETIF_MSG_TX_DONE = 0x0400, 3868 NETIF_MSG_RX_STATUS = 0x0800, 3869 NETIF_MSG_PKTDATA = 0x1000, 3870 NETIF_MSG_HW = 0x2000, 3871 NETIF_MSG_WOL = 0x4000, 3872 }; 3873 3874 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3875 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3876 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3877 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3878 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3879 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3880 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3881 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3882 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3883 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3884 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3885 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3886 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3887 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3888 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3889 3890 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3891 { 3892 /* use default */ 3893 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3894 return default_msg_enable_bits; 3895 if (debug_value == 0) /* no output */ 3896 return 0; 3897 /* set low N bits */ 3898 return (1U << debug_value) - 1; 3899 } 3900 3901 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3902 { 3903 spin_lock(&txq->_xmit_lock); 3904 txq->xmit_lock_owner = cpu; 3905 } 3906 3907 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3908 { 3909 __acquire(&txq->_xmit_lock); 3910 return true; 3911 } 3912 3913 static inline void __netif_tx_release(struct netdev_queue *txq) 3914 { 3915 __release(&txq->_xmit_lock); 3916 } 3917 3918 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3919 { 3920 spin_lock_bh(&txq->_xmit_lock); 3921 txq->xmit_lock_owner = smp_processor_id(); 3922 } 3923 3924 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3925 { 3926 bool ok = spin_trylock(&txq->_xmit_lock); 3927 if (likely(ok)) 3928 txq->xmit_lock_owner = smp_processor_id(); 3929 return ok; 3930 } 3931 3932 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3933 { 3934 txq->xmit_lock_owner = -1; 3935 spin_unlock(&txq->_xmit_lock); 3936 } 3937 3938 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3939 { 3940 txq->xmit_lock_owner = -1; 3941 spin_unlock_bh(&txq->_xmit_lock); 3942 } 3943 3944 static inline void txq_trans_update(struct netdev_queue *txq) 3945 { 3946 if (txq->xmit_lock_owner != -1) 3947 txq->trans_start = jiffies; 3948 } 3949 3950 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3951 static inline void netif_trans_update(struct net_device *dev) 3952 { 3953 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3954 3955 if (txq->trans_start != jiffies) 3956 txq->trans_start = jiffies; 3957 } 3958 3959 /** 3960 * netif_tx_lock - grab network device transmit lock 3961 * @dev: network device 3962 * 3963 * Get network device transmit lock 3964 */ 3965 static inline void netif_tx_lock(struct net_device *dev) 3966 { 3967 unsigned int i; 3968 int cpu; 3969 3970 spin_lock(&dev->tx_global_lock); 3971 cpu = smp_processor_id(); 3972 for (i = 0; i < dev->num_tx_queues; i++) { 3973 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3974 3975 /* We are the only thread of execution doing a 3976 * freeze, but we have to grab the _xmit_lock in 3977 * order to synchronize with threads which are in 3978 * the ->hard_start_xmit() handler and already 3979 * checked the frozen bit. 3980 */ 3981 __netif_tx_lock(txq, cpu); 3982 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3983 __netif_tx_unlock(txq); 3984 } 3985 } 3986 3987 static inline void netif_tx_lock_bh(struct net_device *dev) 3988 { 3989 local_bh_disable(); 3990 netif_tx_lock(dev); 3991 } 3992 3993 static inline void netif_tx_unlock(struct net_device *dev) 3994 { 3995 unsigned int i; 3996 3997 for (i = 0; i < dev->num_tx_queues; i++) { 3998 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3999 4000 /* No need to grab the _xmit_lock here. If the 4001 * queue is not stopped for another reason, we 4002 * force a schedule. 4003 */ 4004 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4005 netif_schedule_queue(txq); 4006 } 4007 spin_unlock(&dev->tx_global_lock); 4008 } 4009 4010 static inline void netif_tx_unlock_bh(struct net_device *dev) 4011 { 4012 netif_tx_unlock(dev); 4013 local_bh_enable(); 4014 } 4015 4016 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4017 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4018 __netif_tx_lock(txq, cpu); \ 4019 } else { \ 4020 __netif_tx_acquire(txq); \ 4021 } \ 4022 } 4023 4024 #define HARD_TX_TRYLOCK(dev, txq) \ 4025 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4026 __netif_tx_trylock(txq) : \ 4027 __netif_tx_acquire(txq)) 4028 4029 #define HARD_TX_UNLOCK(dev, txq) { \ 4030 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4031 __netif_tx_unlock(txq); \ 4032 } else { \ 4033 __netif_tx_release(txq); \ 4034 } \ 4035 } 4036 4037 static inline void netif_tx_disable(struct net_device *dev) 4038 { 4039 unsigned int i; 4040 int cpu; 4041 4042 local_bh_disable(); 4043 cpu = smp_processor_id(); 4044 for (i = 0; i < dev->num_tx_queues; i++) { 4045 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4046 4047 __netif_tx_lock(txq, cpu); 4048 netif_tx_stop_queue(txq); 4049 __netif_tx_unlock(txq); 4050 } 4051 local_bh_enable(); 4052 } 4053 4054 static inline void netif_addr_lock(struct net_device *dev) 4055 { 4056 spin_lock(&dev->addr_list_lock); 4057 } 4058 4059 static inline void netif_addr_lock_nested(struct net_device *dev) 4060 { 4061 int subclass = SINGLE_DEPTH_NESTING; 4062 4063 if (dev->netdev_ops->ndo_get_lock_subclass) 4064 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 4065 4066 spin_lock_nested(&dev->addr_list_lock, subclass); 4067 } 4068 4069 static inline void netif_addr_lock_bh(struct net_device *dev) 4070 { 4071 spin_lock_bh(&dev->addr_list_lock); 4072 } 4073 4074 static inline void netif_addr_unlock(struct net_device *dev) 4075 { 4076 spin_unlock(&dev->addr_list_lock); 4077 } 4078 4079 static inline void netif_addr_unlock_bh(struct net_device *dev) 4080 { 4081 spin_unlock_bh(&dev->addr_list_lock); 4082 } 4083 4084 /* 4085 * dev_addrs walker. Should be used only for read access. Call with 4086 * rcu_read_lock held. 4087 */ 4088 #define for_each_dev_addr(dev, ha) \ 4089 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4090 4091 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4092 4093 void ether_setup(struct net_device *dev); 4094 4095 /* Support for loadable net-drivers */ 4096 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4097 unsigned char name_assign_type, 4098 void (*setup)(struct net_device *), 4099 unsigned int txqs, unsigned int rxqs); 4100 int dev_get_valid_name(struct net *net, struct net_device *dev, 4101 const char *name); 4102 4103 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4104 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4105 4106 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4107 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4108 count) 4109 4110 int register_netdev(struct net_device *dev); 4111 void unregister_netdev(struct net_device *dev); 4112 4113 /* General hardware address lists handling functions */ 4114 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4115 struct netdev_hw_addr_list *from_list, int addr_len); 4116 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4117 struct netdev_hw_addr_list *from_list, int addr_len); 4118 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4119 struct net_device *dev, 4120 int (*sync)(struct net_device *, const unsigned char *), 4121 int (*unsync)(struct net_device *, 4122 const unsigned char *)); 4123 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4124 struct net_device *dev, 4125 int (*sync)(struct net_device *, 4126 const unsigned char *, int), 4127 int (*unsync)(struct net_device *, 4128 const unsigned char *, int)); 4129 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4130 struct net_device *dev, 4131 int (*unsync)(struct net_device *, 4132 const unsigned char *, int)); 4133 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4134 struct net_device *dev, 4135 int (*unsync)(struct net_device *, 4136 const unsigned char *)); 4137 void __hw_addr_init(struct netdev_hw_addr_list *list); 4138 4139 /* Functions used for device addresses handling */ 4140 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4141 unsigned char addr_type); 4142 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4143 unsigned char addr_type); 4144 void dev_addr_flush(struct net_device *dev); 4145 int dev_addr_init(struct net_device *dev); 4146 4147 /* Functions used for unicast addresses handling */ 4148 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4149 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4150 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4151 int dev_uc_sync(struct net_device *to, struct net_device *from); 4152 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4153 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4154 void dev_uc_flush(struct net_device *dev); 4155 void dev_uc_init(struct net_device *dev); 4156 4157 /** 4158 * __dev_uc_sync - Synchonize device's unicast list 4159 * @dev: device to sync 4160 * @sync: function to call if address should be added 4161 * @unsync: function to call if address should be removed 4162 * 4163 * Add newly added addresses to the interface, and release 4164 * addresses that have been deleted. 4165 */ 4166 static inline int __dev_uc_sync(struct net_device *dev, 4167 int (*sync)(struct net_device *, 4168 const unsigned char *), 4169 int (*unsync)(struct net_device *, 4170 const unsigned char *)) 4171 { 4172 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4173 } 4174 4175 /** 4176 * __dev_uc_unsync - Remove synchronized addresses from device 4177 * @dev: device to sync 4178 * @unsync: function to call if address should be removed 4179 * 4180 * Remove all addresses that were added to the device by dev_uc_sync(). 4181 */ 4182 static inline void __dev_uc_unsync(struct net_device *dev, 4183 int (*unsync)(struct net_device *, 4184 const unsigned char *)) 4185 { 4186 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4187 } 4188 4189 /* Functions used for multicast addresses handling */ 4190 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4191 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4192 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4193 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4194 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4195 int dev_mc_sync(struct net_device *to, struct net_device *from); 4196 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4197 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4198 void dev_mc_flush(struct net_device *dev); 4199 void dev_mc_init(struct net_device *dev); 4200 4201 /** 4202 * __dev_mc_sync - Synchonize device's multicast list 4203 * @dev: device to sync 4204 * @sync: function to call if address should be added 4205 * @unsync: function to call if address should be removed 4206 * 4207 * Add newly added addresses to the interface, and release 4208 * addresses that have been deleted. 4209 */ 4210 static inline int __dev_mc_sync(struct net_device *dev, 4211 int (*sync)(struct net_device *, 4212 const unsigned char *), 4213 int (*unsync)(struct net_device *, 4214 const unsigned char *)) 4215 { 4216 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4217 } 4218 4219 /** 4220 * __dev_mc_unsync - Remove synchronized addresses from device 4221 * @dev: device to sync 4222 * @unsync: function to call if address should be removed 4223 * 4224 * Remove all addresses that were added to the device by dev_mc_sync(). 4225 */ 4226 static inline void __dev_mc_unsync(struct net_device *dev, 4227 int (*unsync)(struct net_device *, 4228 const unsigned char *)) 4229 { 4230 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4231 } 4232 4233 /* Functions used for secondary unicast and multicast support */ 4234 void dev_set_rx_mode(struct net_device *dev); 4235 void __dev_set_rx_mode(struct net_device *dev); 4236 int dev_set_promiscuity(struct net_device *dev, int inc); 4237 int dev_set_allmulti(struct net_device *dev, int inc); 4238 void netdev_state_change(struct net_device *dev); 4239 void netdev_notify_peers(struct net_device *dev); 4240 void netdev_features_change(struct net_device *dev); 4241 /* Load a device via the kmod */ 4242 void dev_load(struct net *net, const char *name); 4243 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4244 struct rtnl_link_stats64 *storage); 4245 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4246 const struct net_device_stats *netdev_stats); 4247 4248 extern int netdev_max_backlog; 4249 extern int netdev_tstamp_prequeue; 4250 extern int weight_p; 4251 extern int dev_weight_rx_bias; 4252 extern int dev_weight_tx_bias; 4253 extern int dev_rx_weight; 4254 extern int dev_tx_weight; 4255 extern int gro_normal_batch; 4256 4257 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4258 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4259 struct list_head **iter); 4260 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4261 struct list_head **iter); 4262 4263 /* iterate through upper list, must be called under RCU read lock */ 4264 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4265 for (iter = &(dev)->adj_list.upper, \ 4266 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4267 updev; \ 4268 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4269 4270 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4271 int (*fn)(struct net_device *upper_dev, 4272 void *data), 4273 void *data); 4274 4275 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4276 struct net_device *upper_dev); 4277 4278 bool netdev_has_any_upper_dev(struct net_device *dev); 4279 4280 void *netdev_lower_get_next_private(struct net_device *dev, 4281 struct list_head **iter); 4282 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4283 struct list_head **iter); 4284 4285 #define netdev_for_each_lower_private(dev, priv, iter) \ 4286 for (iter = (dev)->adj_list.lower.next, \ 4287 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4288 priv; \ 4289 priv = netdev_lower_get_next_private(dev, &(iter))) 4290 4291 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4292 for (iter = &(dev)->adj_list.lower, \ 4293 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4294 priv; \ 4295 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4296 4297 void *netdev_lower_get_next(struct net_device *dev, 4298 struct list_head **iter); 4299 4300 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4301 for (iter = (dev)->adj_list.lower.next, \ 4302 ldev = netdev_lower_get_next(dev, &(iter)); \ 4303 ldev; \ 4304 ldev = netdev_lower_get_next(dev, &(iter))) 4305 4306 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 4307 struct list_head **iter); 4308 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 4309 struct list_head **iter); 4310 4311 int netdev_walk_all_lower_dev(struct net_device *dev, 4312 int (*fn)(struct net_device *lower_dev, 4313 void *data), 4314 void *data); 4315 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4316 int (*fn)(struct net_device *lower_dev, 4317 void *data), 4318 void *data); 4319 4320 void *netdev_adjacent_get_private(struct list_head *adj_list); 4321 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4322 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4323 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4324 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4325 struct netlink_ext_ack *extack); 4326 int netdev_master_upper_dev_link(struct net_device *dev, 4327 struct net_device *upper_dev, 4328 void *upper_priv, void *upper_info, 4329 struct netlink_ext_ack *extack); 4330 void netdev_upper_dev_unlink(struct net_device *dev, 4331 struct net_device *upper_dev); 4332 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4333 void *netdev_lower_dev_get_private(struct net_device *dev, 4334 struct net_device *lower_dev); 4335 void netdev_lower_state_changed(struct net_device *lower_dev, 4336 void *lower_state_info); 4337 4338 /* RSS keys are 40 or 52 bytes long */ 4339 #define NETDEV_RSS_KEY_LEN 52 4340 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4341 void netdev_rss_key_fill(void *buffer, size_t len); 4342 4343 int dev_get_nest_level(struct net_device *dev); 4344 int skb_checksum_help(struct sk_buff *skb); 4345 int skb_crc32c_csum_help(struct sk_buff *skb); 4346 int skb_csum_hwoffload_help(struct sk_buff *skb, 4347 const netdev_features_t features); 4348 4349 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4350 netdev_features_t features, bool tx_path); 4351 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4352 netdev_features_t features); 4353 4354 struct netdev_bonding_info { 4355 ifslave slave; 4356 ifbond master; 4357 }; 4358 4359 struct netdev_notifier_bonding_info { 4360 struct netdev_notifier_info info; /* must be first */ 4361 struct netdev_bonding_info bonding_info; 4362 }; 4363 4364 void netdev_bonding_info_change(struct net_device *dev, 4365 struct netdev_bonding_info *bonding_info); 4366 4367 static inline 4368 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4369 { 4370 return __skb_gso_segment(skb, features, true); 4371 } 4372 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4373 4374 static inline bool can_checksum_protocol(netdev_features_t features, 4375 __be16 protocol) 4376 { 4377 if (protocol == htons(ETH_P_FCOE)) 4378 return !!(features & NETIF_F_FCOE_CRC); 4379 4380 /* Assume this is an IP checksum (not SCTP CRC) */ 4381 4382 if (features & NETIF_F_HW_CSUM) { 4383 /* Can checksum everything */ 4384 return true; 4385 } 4386 4387 switch (protocol) { 4388 case htons(ETH_P_IP): 4389 return !!(features & NETIF_F_IP_CSUM); 4390 case htons(ETH_P_IPV6): 4391 return !!(features & NETIF_F_IPV6_CSUM); 4392 default: 4393 return false; 4394 } 4395 } 4396 4397 #ifdef CONFIG_BUG 4398 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4399 #else 4400 static inline void netdev_rx_csum_fault(struct net_device *dev, 4401 struct sk_buff *skb) 4402 { 4403 } 4404 #endif 4405 /* rx skb timestamps */ 4406 void net_enable_timestamp(void); 4407 void net_disable_timestamp(void); 4408 4409 #ifdef CONFIG_PROC_FS 4410 int __init dev_proc_init(void); 4411 #else 4412 #define dev_proc_init() 0 4413 #endif 4414 4415 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4416 struct sk_buff *skb, struct net_device *dev, 4417 bool more) 4418 { 4419 __this_cpu_write(softnet_data.xmit.more, more); 4420 return ops->ndo_start_xmit(skb, dev); 4421 } 4422 4423 static inline bool netdev_xmit_more(void) 4424 { 4425 return __this_cpu_read(softnet_data.xmit.more); 4426 } 4427 4428 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4429 struct netdev_queue *txq, bool more) 4430 { 4431 const struct net_device_ops *ops = dev->netdev_ops; 4432 netdev_tx_t rc; 4433 4434 rc = __netdev_start_xmit(ops, skb, dev, more); 4435 if (rc == NETDEV_TX_OK) 4436 txq_trans_update(txq); 4437 4438 return rc; 4439 } 4440 4441 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4442 const void *ns); 4443 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4444 const void *ns); 4445 4446 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4447 { 4448 return netdev_class_create_file_ns(class_attr, NULL); 4449 } 4450 4451 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4452 { 4453 netdev_class_remove_file_ns(class_attr, NULL); 4454 } 4455 4456 extern const struct kobj_ns_type_operations net_ns_type_operations; 4457 4458 const char *netdev_drivername(const struct net_device *dev); 4459 4460 void linkwatch_run_queue(void); 4461 4462 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4463 netdev_features_t f2) 4464 { 4465 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4466 if (f1 & NETIF_F_HW_CSUM) 4467 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4468 else 4469 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4470 } 4471 4472 return f1 & f2; 4473 } 4474 4475 static inline netdev_features_t netdev_get_wanted_features( 4476 struct net_device *dev) 4477 { 4478 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4479 } 4480 netdev_features_t netdev_increment_features(netdev_features_t all, 4481 netdev_features_t one, netdev_features_t mask); 4482 4483 /* Allow TSO being used on stacked device : 4484 * Performing the GSO segmentation before last device 4485 * is a performance improvement. 4486 */ 4487 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4488 netdev_features_t mask) 4489 { 4490 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4491 } 4492 4493 int __netdev_update_features(struct net_device *dev); 4494 void netdev_update_features(struct net_device *dev); 4495 void netdev_change_features(struct net_device *dev); 4496 4497 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4498 struct net_device *dev); 4499 4500 netdev_features_t passthru_features_check(struct sk_buff *skb, 4501 struct net_device *dev, 4502 netdev_features_t features); 4503 netdev_features_t netif_skb_features(struct sk_buff *skb); 4504 4505 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4506 { 4507 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4508 4509 /* check flags correspondence */ 4510 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4511 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4512 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4513 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4514 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4515 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4516 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4517 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4518 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4519 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4520 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4521 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4522 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4523 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4524 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4525 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4526 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4527 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4528 4529 return (features & feature) == feature; 4530 } 4531 4532 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4533 { 4534 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4535 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4536 } 4537 4538 static inline bool netif_needs_gso(struct sk_buff *skb, 4539 netdev_features_t features) 4540 { 4541 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4542 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4543 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4544 } 4545 4546 static inline void netif_set_gso_max_size(struct net_device *dev, 4547 unsigned int size) 4548 { 4549 dev->gso_max_size = size; 4550 } 4551 4552 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4553 int pulled_hlen, u16 mac_offset, 4554 int mac_len) 4555 { 4556 skb->protocol = protocol; 4557 skb->encapsulation = 1; 4558 skb_push(skb, pulled_hlen); 4559 skb_reset_transport_header(skb); 4560 skb->mac_header = mac_offset; 4561 skb->network_header = skb->mac_header + mac_len; 4562 skb->mac_len = mac_len; 4563 } 4564 4565 static inline bool netif_is_macsec(const struct net_device *dev) 4566 { 4567 return dev->priv_flags & IFF_MACSEC; 4568 } 4569 4570 static inline bool netif_is_macvlan(const struct net_device *dev) 4571 { 4572 return dev->priv_flags & IFF_MACVLAN; 4573 } 4574 4575 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4576 { 4577 return dev->priv_flags & IFF_MACVLAN_PORT; 4578 } 4579 4580 static inline bool netif_is_bond_master(const struct net_device *dev) 4581 { 4582 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4583 } 4584 4585 static inline bool netif_is_bond_slave(const struct net_device *dev) 4586 { 4587 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4588 } 4589 4590 static inline bool netif_supports_nofcs(struct net_device *dev) 4591 { 4592 return dev->priv_flags & IFF_SUPP_NOFCS; 4593 } 4594 4595 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4596 { 4597 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4598 } 4599 4600 static inline bool netif_is_l3_master(const struct net_device *dev) 4601 { 4602 return dev->priv_flags & IFF_L3MDEV_MASTER; 4603 } 4604 4605 static inline bool netif_is_l3_slave(const struct net_device *dev) 4606 { 4607 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4608 } 4609 4610 static inline bool netif_is_bridge_master(const struct net_device *dev) 4611 { 4612 return dev->priv_flags & IFF_EBRIDGE; 4613 } 4614 4615 static inline bool netif_is_bridge_port(const struct net_device *dev) 4616 { 4617 return dev->priv_flags & IFF_BRIDGE_PORT; 4618 } 4619 4620 static inline bool netif_is_ovs_master(const struct net_device *dev) 4621 { 4622 return dev->priv_flags & IFF_OPENVSWITCH; 4623 } 4624 4625 static inline bool netif_is_ovs_port(const struct net_device *dev) 4626 { 4627 return dev->priv_flags & IFF_OVS_DATAPATH; 4628 } 4629 4630 static inline bool netif_is_team_master(const struct net_device *dev) 4631 { 4632 return dev->priv_flags & IFF_TEAM; 4633 } 4634 4635 static inline bool netif_is_team_port(const struct net_device *dev) 4636 { 4637 return dev->priv_flags & IFF_TEAM_PORT; 4638 } 4639 4640 static inline bool netif_is_lag_master(const struct net_device *dev) 4641 { 4642 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4643 } 4644 4645 static inline bool netif_is_lag_port(const struct net_device *dev) 4646 { 4647 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4648 } 4649 4650 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4651 { 4652 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4653 } 4654 4655 static inline bool netif_is_failover(const struct net_device *dev) 4656 { 4657 return dev->priv_flags & IFF_FAILOVER; 4658 } 4659 4660 static inline bool netif_is_failover_slave(const struct net_device *dev) 4661 { 4662 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4663 } 4664 4665 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4666 static inline void netif_keep_dst(struct net_device *dev) 4667 { 4668 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4669 } 4670 4671 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4672 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4673 { 4674 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4675 return dev->priv_flags & IFF_MACSEC; 4676 } 4677 4678 extern struct pernet_operations __net_initdata loopback_net_ops; 4679 4680 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4681 4682 /* netdev_printk helpers, similar to dev_printk */ 4683 4684 static inline const char *netdev_name(const struct net_device *dev) 4685 { 4686 if (!dev->name[0] || strchr(dev->name, '%')) 4687 return "(unnamed net_device)"; 4688 return dev->name; 4689 } 4690 4691 static inline bool netdev_unregistering(const struct net_device *dev) 4692 { 4693 return dev->reg_state == NETREG_UNREGISTERING; 4694 } 4695 4696 static inline const char *netdev_reg_state(const struct net_device *dev) 4697 { 4698 switch (dev->reg_state) { 4699 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4700 case NETREG_REGISTERED: return ""; 4701 case NETREG_UNREGISTERING: return " (unregistering)"; 4702 case NETREG_UNREGISTERED: return " (unregistered)"; 4703 case NETREG_RELEASED: return " (released)"; 4704 case NETREG_DUMMY: return " (dummy)"; 4705 } 4706 4707 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4708 return " (unknown)"; 4709 } 4710 4711 __printf(3, 4) __cold 4712 void netdev_printk(const char *level, const struct net_device *dev, 4713 const char *format, ...); 4714 __printf(2, 3) __cold 4715 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4716 __printf(2, 3) __cold 4717 void netdev_alert(const struct net_device *dev, const char *format, ...); 4718 __printf(2, 3) __cold 4719 void netdev_crit(const struct net_device *dev, const char *format, ...); 4720 __printf(2, 3) __cold 4721 void netdev_err(const struct net_device *dev, const char *format, ...); 4722 __printf(2, 3) __cold 4723 void netdev_warn(const struct net_device *dev, const char *format, ...); 4724 __printf(2, 3) __cold 4725 void netdev_notice(const struct net_device *dev, const char *format, ...); 4726 __printf(2, 3) __cold 4727 void netdev_info(const struct net_device *dev, const char *format, ...); 4728 4729 #define netdev_level_once(level, dev, fmt, ...) \ 4730 do { \ 4731 static bool __print_once __read_mostly; \ 4732 \ 4733 if (!__print_once) { \ 4734 __print_once = true; \ 4735 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4736 } \ 4737 } while (0) 4738 4739 #define netdev_emerg_once(dev, fmt, ...) \ 4740 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4741 #define netdev_alert_once(dev, fmt, ...) \ 4742 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4743 #define netdev_crit_once(dev, fmt, ...) \ 4744 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4745 #define netdev_err_once(dev, fmt, ...) \ 4746 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4747 #define netdev_warn_once(dev, fmt, ...) \ 4748 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4749 #define netdev_notice_once(dev, fmt, ...) \ 4750 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4751 #define netdev_info_once(dev, fmt, ...) \ 4752 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4753 4754 #define MODULE_ALIAS_NETDEV(device) \ 4755 MODULE_ALIAS("netdev-" device) 4756 4757 #if defined(CONFIG_DYNAMIC_DEBUG) 4758 #define netdev_dbg(__dev, format, args...) \ 4759 do { \ 4760 dynamic_netdev_dbg(__dev, format, ##args); \ 4761 } while (0) 4762 #elif defined(DEBUG) 4763 #define netdev_dbg(__dev, format, args...) \ 4764 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4765 #else 4766 #define netdev_dbg(__dev, format, args...) \ 4767 ({ \ 4768 if (0) \ 4769 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4770 }) 4771 #endif 4772 4773 #if defined(VERBOSE_DEBUG) 4774 #define netdev_vdbg netdev_dbg 4775 #else 4776 4777 #define netdev_vdbg(dev, format, args...) \ 4778 ({ \ 4779 if (0) \ 4780 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4781 0; \ 4782 }) 4783 #endif 4784 4785 /* 4786 * netdev_WARN() acts like dev_printk(), but with the key difference 4787 * of using a WARN/WARN_ON to get the message out, including the 4788 * file/line information and a backtrace. 4789 */ 4790 #define netdev_WARN(dev, format, args...) \ 4791 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4792 netdev_reg_state(dev), ##args) 4793 4794 #define netdev_WARN_ONCE(dev, format, args...) \ 4795 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4796 netdev_reg_state(dev), ##args) 4797 4798 /* netif printk helpers, similar to netdev_printk */ 4799 4800 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4801 do { \ 4802 if (netif_msg_##type(priv)) \ 4803 netdev_printk(level, (dev), fmt, ##args); \ 4804 } while (0) 4805 4806 #define netif_level(level, priv, type, dev, fmt, args...) \ 4807 do { \ 4808 if (netif_msg_##type(priv)) \ 4809 netdev_##level(dev, fmt, ##args); \ 4810 } while (0) 4811 4812 #define netif_emerg(priv, type, dev, fmt, args...) \ 4813 netif_level(emerg, priv, type, dev, fmt, ##args) 4814 #define netif_alert(priv, type, dev, fmt, args...) \ 4815 netif_level(alert, priv, type, dev, fmt, ##args) 4816 #define netif_crit(priv, type, dev, fmt, args...) \ 4817 netif_level(crit, priv, type, dev, fmt, ##args) 4818 #define netif_err(priv, type, dev, fmt, args...) \ 4819 netif_level(err, priv, type, dev, fmt, ##args) 4820 #define netif_warn(priv, type, dev, fmt, args...) \ 4821 netif_level(warn, priv, type, dev, fmt, ##args) 4822 #define netif_notice(priv, type, dev, fmt, args...) \ 4823 netif_level(notice, priv, type, dev, fmt, ##args) 4824 #define netif_info(priv, type, dev, fmt, args...) \ 4825 netif_level(info, priv, type, dev, fmt, ##args) 4826 4827 #if defined(CONFIG_DYNAMIC_DEBUG) 4828 #define netif_dbg(priv, type, netdev, format, args...) \ 4829 do { \ 4830 if (netif_msg_##type(priv)) \ 4831 dynamic_netdev_dbg(netdev, format, ##args); \ 4832 } while (0) 4833 #elif defined(DEBUG) 4834 #define netif_dbg(priv, type, dev, format, args...) \ 4835 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4836 #else 4837 #define netif_dbg(priv, type, dev, format, args...) \ 4838 ({ \ 4839 if (0) \ 4840 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4841 0; \ 4842 }) 4843 #endif 4844 4845 /* if @cond then downgrade to debug, else print at @level */ 4846 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4847 do { \ 4848 if (cond) \ 4849 netif_dbg(priv, type, netdev, fmt, ##args); \ 4850 else \ 4851 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4852 } while (0) 4853 4854 #if defined(VERBOSE_DEBUG) 4855 #define netif_vdbg netif_dbg 4856 #else 4857 #define netif_vdbg(priv, type, dev, format, args...) \ 4858 ({ \ 4859 if (0) \ 4860 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4861 0; \ 4862 }) 4863 #endif 4864 4865 /* 4866 * The list of packet types we will receive (as opposed to discard) 4867 * and the routines to invoke. 4868 * 4869 * Why 16. Because with 16 the only overlap we get on a hash of the 4870 * low nibble of the protocol value is RARP/SNAP/X.25. 4871 * 4872 * 0800 IP 4873 * 0001 802.3 4874 * 0002 AX.25 4875 * 0004 802.2 4876 * 8035 RARP 4877 * 0005 SNAP 4878 * 0805 X.25 4879 * 0806 ARP 4880 * 8137 IPX 4881 * 0009 Localtalk 4882 * 86DD IPv6 4883 */ 4884 #define PTYPE_HASH_SIZE (16) 4885 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4886 4887 extern struct net_device *blackhole_netdev; 4888 4889 #endif /* _LINUX_NETDEVICE_H */ 4890